MIDREAL
Menu

Gemini 4 Neon

https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/ (blog.google)

1037 points | 3d ago | Discuss on Hacker News | Back to Radar

Comments

brap 3d ago on HN
Google is on a roll, what is this the 4th release this month?
nateb2022 3d ago on HN
RSI
brap 3d ago on HN
Sure looks like it
catigula 3d ago on HN
It looks like a distilled model.
inknight 3d ago on HN
Please have them recursively improve their model names
mpvjs 3d ago on HN
- please improve the model names

ChatGPT thinking: the user is not happy with my previous solution, propose a cooler scheme like toy story character names...

tanx16 3d ago on HN
They're releasing... periodically
bloky 3d ago on HN
At this point you just use roulette on which Gemini model you'll use for a task
BoldColdHold 3d ago on HN
Ok this is a good model by the looks of it. Mid size model that only loses to fable.
WarmWash 3d ago on HN
Neon is the long awaited successor to 3.1 pro. Took a while to get there

So now we have

(XL): Carbon/Argon Astra Fable

(L): Neon Sol Opus

(M): Hydrogen Terra Sonnet

(S): Helix(rumored) Luna Haiku

5555watch 3d ago on HN
I don't believe Argon is XL level??

Also, very curious whether they will ever release the rumored Krypton model, or will it be hidden ala Mythos

WarmWash 3d ago on HN
Argon is a big model, but it was completed sometime in mid-august, making it kind of old in LLM land.
Melatonic 3d ago on HN
Other people are saying that Neon replaced Argon?
WarmWash 3d ago on HN
It very well might be better. Nowadays being newer tends to matter more than being bigger.
agentdev001 3d ago on HN
Yea- if we're just talking size these would be the tiers, but if you're measuring by API spend then things certainly get messy.
oblio 3d ago on HN
> Neon is the long awaited successor to 3.1 pro. Took a while to get there.

The real question is: will it be available for Google Workspace? Not everyone wants to pay for a dedicated AI license :-(

mrbungie 3d ago on HN
Is this another release that I can't use with agy? (I totally refuse to pay 100 USD just to try it in that context)

Edit: I stand corrected, it can be used in agy.

Edit2: I have an student account and students accounts can't use these models apparently, https://discuss.ai.google.dev/t/bug-entitlement-sync-paid-go...

lanthissa 3d ago on HN
you can use this with agy, im on pro and just tried it it has its own orchestrator too for delegating to hydrogen.
mrbungie 3d ago on HN
Oh cool, I'll check it then. Last time I checked I only had 3.8 flash and 3.1 pro.
anvuong 3d ago on HN
It's there in my $20 sub. Only Carbon needs $100+
3wolf 3d ago on HN
Credit to Google for outdoing Open AI and Anthropic in terms of choosing a confusing naming convention.
xattt 3d ago on HN
Is it not going by atomic mass?
daemonologist 3d ago on HN
Only if you discount Argon (~40)
XCSme 3d ago on HN
So it's better and cheaper than the recently released Carbon?

So Carbon is already deprecated, or?

EDIT: They do explain where it sits in the line-up here:

https://blog.google/innovation-and-ai/models-and-research/ge...

jrflo 3d ago on HN
Between Astra and Fable on the Pareto frontier:

https://artificialanalysis.ai/#intelligence-efficiency-tabs

jeffbee 3d ago on HN
That looks to me like OpenAI equals or beats everything Google offers at half the price.
jrflo 3d ago on HN
You're right, it's 1-2 points shy of the frontier
jeffbee 3d ago on HN
I'm surprised they can't afford to just slide the prices over to the left and eat the losses.
scotty79 3d ago on HN
API prices are arbitrarily set though. Google set their pricing to control the demand rather than because it had to be that high.

I heard that Google share price doesn't really depend on how much their models are used. But rather how do the bench and whether they are theoretically publicly accessible so that their quality can be independantly confirmed. So Google can price their model higher to get most of the benefits without having to actually provide too much inference.

They are probably releasing the models to the public at all only because they got so much backlash after witholding Argon.

nicoburns 3d ago on HN
Is there a way to determine "model size class" from Google's naming scheme?
catigula 3d ago on HN
If there is, it isn’t obvious, which makes it terrible.
XCSme 3d ago on HN
jeffbee 3d ago on HN
You violated the only constraint the question contained.
XCSme 3d ago on HN
Think of it like a riddle: given the order and naming scheme, OP has to imagine what Google thinks when naming the models...
ehsankia 3d ago on HN
Honestly, the bigger question is, are they at some point start adding numbers (Carbon 2, Neon 1.5, Hydrogen+ ionized), or are they keep giving a new name to every single new model. Argon seems to have been basically "replaced" by Carbon.
kzrdude 3d ago on HN
I hope they will continue to use numbers. Right now it's "Gemini 4 Hydrogen" in the UI, that seems to be the way it will be used.
lukax 3d ago on HN
Carbon is just Argon with extended thinking.
andai 3d ago on HN
Subscribing for Gemini 19 Bismuth
grantsucceeded 3d ago on HN
definitely the funnniest of all element names :)
antonvs 3d ago on HN
Looking up the model name in a table is "a way to determine 'model size class' from Google's naming scheme".
troupo 3d ago on HN
Leave it to Google to create a card that goes from highest to lowest, and not the other way around
onlyj 3d ago on HN
Why does this chart not include the other gemini 4 model, Argon? Also I am sure somebody thought about which element to name which model and it would be very helpful to me if they would explain their reasoning. OpenAI's naming got kind of abstract with 5.6 but I could follow it: Luna(Moon) was smallest, Terra (Earth) was medium, Sol (Sun) was biggest. Astra kind of breaks from that but whatever. But like I would genuinely appreciation someone from google to explain the reasoning behind their element names. Like are Argon and Neon related in some way because they are both noble gases? Is hydrogen the smallest model because it is the lightest element? Carbon has a lower atomic number than neon but is more expensive and presumably bigger... maybe I am overthinking this.
thevillagechief 3d ago on HN
Argon is already deprecated.
_delirium 3d ago on HN
Inconsistently it seems. On the $20/mo tier, Argon has been removed from the web interface and app, but is still available in Antigravity (while Carbon is only on the $100/mo tier in both).
jeffbee 3d ago on HN
Uh yeah, is this the big news of this announcement? They are moving all the good coding models into the Ultra price tier.
Ternari 3d ago on HN
Are they? Neon seems great and it’s available on the Pro tier.
jeffbee 3d ago on HN
I'm going by the model labels in Gemini.
LewisOrton 3d ago on HN
is it supposed to be replaced by neon?
drum55 3d ago on HN
The developer guide lists

    Gemini 4 Neon (NEW!)
    Gemini 4 Hydrogen (NEW!)
    Gemini 4 Carbon (NEW!)
    Gemini 4 Argon (NEW!
    Gemini 3.8 Live 
    Gemini 3.8 Flash (NEW!)
    Gemini 3.7 Flash
    Gemini 3.6 Flash
    Gemini 3.5 Flash
I've no clue whatsoever how to read anything about the names of them and what they might imply you would use them for though. Within these most of them have LOW MEDIUM HIGH XHIGH, with the default being different for some of them, how anybody is possibly meant to understand this whole matrix of ambiguously named products is beyond me.
aenis 3d ago on HN
Same here. First i thought it would be noble gasses. Then I thought it would be about the atomic weights. I know, I should just ask AI to figure it out for me, but its a bit humiliating.
drum55 3d ago on HN

    Hydrogen 1
    Carbon 6 
    Neon 10
    Argon 18
Maybe this implies that Neon is smaller than Argon in weight size, but it's newer, but it's also only a little bigger than Carbon? Do the Nobel gas models have more safeguards? Is Carbon 6x the size of Hydrogen and a third the size of Argon? When they release Xeon can we make compounds? Feels like some really cheap Da Vinci Code esque thing if we're supposed to get some sort of meaning out of the names.
svnt 3d ago on HN
From their Neon marketing:

> In our early evaluations, it matched or exceeded Gemini 4 Carbon on overall quality at 3x the speed and 60% of the cost. We’re excited to bring these capabilities to Glean customers tackling multi-step work that demands accuracy, detail, and speed.

So possibly smaller than Carbon? Or some other axis that is making it faster and cheaper? No idea.

aenis 3d ago on HN
Same. Running my benchmarks to verify their quality claims. Carbon performed very well on my private suite - spitting distance to Opus 5.5.
aenis 3d ago on HN
...first runs are in, and Neon is far below Carbon in terms of quality, but indeed much faster - ~4x faster, and 2x cheaper per task. Probably a good model for delegating to after key design work is done with something cleverer.
svnt 3d ago on HN
What category of work did it underperform on for you? I am seeing comparable outcomes in some technical document processing, but the cost is not lower for me.
aenis 3d ago on HN
Gold plating for tasks where the agent is given the intention, the explicit ask to make the best impression, and ways of self evaluating. Much like openai's agents pre 6.2 sol/astra, it considers the task satisfactorily completed a bit earlier than anthropic models which seem to better guess what the specs left out. It is a useless feature for those who keep their llms at a short leash - my hobby benchmark aims to evaluate performance at under-specified tasks and, well, vibe coding as it was described a year ago.

It's a good model for delegating work to, but not deciding the detailed scope. So, not fable class for me.

sheepscreek 2d ago on HN
Sundar Pichai (must have) himself signed off on the names.
iainmerrick 3d ago on HN
It could be worse, they could have named them after different Google chat products.
jimbokun 3d ago on HN
How do you decide which model is best for telling you which model to use?
prymitive 3d ago on HN
It’s just a matter of time before we have

- Gemini (New)

- Gemini

- Gemini (original)

abixb 3d ago on HN
Inb4 they switch it back to Bard.
accrual 3d ago on HN
They could take a page from Nintendo: New Super Gemini Deluxe!
murderfs 3d ago on HN
This is Google, it'd be:

- Gemini (Not ready yet)

- Gemini (Deprecated)

jimbokun 3d ago on HN
Classic Gemini
kzrdude 3d ago on HN
Hydrogen is the new flash, at least
elxr 3d ago on HN
Don't use Argon or Carbon.

Hydrogen is the cheap one for straightforward tasks.

Neon is supposedly Opus-level, which is the exciting one if it actually works as well as they claim.

Obviously anything with a 3.x is not what you should reach for. This has always been the case on a first digit bump in any of these mainstream models.

genxy 3d ago on HN
I think you mean say nobel size, I'd wait for radon.
swasheck 3d ago on HN
now we have to know our atomic weights to be able to choose which gemini model for the task, i suppose.
CamperBob2 3d ago on HN
Problem with Radon is that they will swap it out for Polonium without telling you, and then they will obsolete that in favor of Lead.
moonu 3d ago on HN
melting point/boiling point
ahofmann 3d ago on HN
Ugh, this sound so overly nerdy, that it could be very true. I do not follow or care about the Google LLM releases just because the naming scheme is so convoluted (and the constant churn of products around the llms). I learned the anthropic naming scheme, it is stable and easy to follow. And this is why Claude is my daily driver. I don't like change, I like stable and predictable progress. Open AI naming scheme is kind of easy to learn (bigger planet = bigger model). But fuck Googles bullshit naming crap, seriously.
onlyrealcuzzo 3d ago on HN
I suspect they may want to get away from the size class due to the fact that - at the pace they are launching new models - and the similarity of model sizes between Neon/Carbon, Opus/Fable, Sol/Astra - the best model is going to be the most recently released one, not the biggest one.

I'm not sure it's much better than 4.1 Flash vs 4 Pro -> if 4.1 Flash is better temporarily when Pro is typically significantly better, people are going to be confused.

I'm not really sure how you get around this problem easily.

It's not going to age well when the next Flash-lite model launches, but maybe they will call that `Gemini 4 <Element>-lite` - who knows...

nicoburns 3d ago on HN
Interesting. You're claiming that the Opus/Sol/Neon class and Fable/Astra/Carbon class models are similar sizes to each other? That would be surprising to me.

I've found that while Opus 5.5 is often pretty nice for practical work, Fable (even 5.0) is still significantly smarter on hard tasks and makes far fewer "dumb mistakes"

dannyw 3d ago on HN
A Microsoft Azure page briefly claimed Sol runs with less looped transformer passes; and so you could speculate that Sol and Astra might share weights; but have different “loop” factors.

That would make a lot of sense from an inference optimisation perspective. The same prefill and decode machines can hold one set of weights, and serve both Astra and Sol requests.

At (multi-)terabyte-scale model weights, and when you have to offer not just the latest like Fable 5.2 but still serve 5.1, 5.0, etc; heck I think Opus 4.6 is still around; you really care about overall compute utilisation.

Idle GPUs are wasted GPUs; and even over the fastest interlinks you’re not hotswapping multi terabytes of weights.

With MoEs, there’s probably stuff you can play around with varying the number of active experts; or other clever optimisations.

Maybe Sonnet 5.5 could be just Opus 5.5 with a different router and less active params? Who knows; I’m just theorycrafting now :)

nicoburns 3d ago on HN
I heard rumours that Sol 6.1 was cut down Astra because Sol 6.0 was received badly. But I thought that was the exception (and a result of Open AI's actual Sol-class model being uncompetitive) rather than the rule.
willy_k 3d ago on HN
That would line up - on release and for the few following weeks Sol 6.1 felt like Astra but maybe more task oriented, recently it’s been a noticeable downgrade in the level of specification needed to successfully set off.
onlyrealcuzzo 3d ago on HN
> A Microsoft Azure page briefly claimed Sol runs with less looped transformer passes;

This doesn't really make sense given how the models released.

OpenAI released Astra long before latent-space looping was a thing at the frontier space.

Then they released Sol 6, which was fast, clearly not looped. Then they release Sol 6.1 which was nearly as good as Astra, but way cheaper, and WAY slower -> this would follow it just being a looped version of Sol 6.

It doesn't seem likely that Astra 6.2 is just a looped Astra 6... It's too fast, and they'd go bankrupt if they're charging the same price... It may be a further looped version of Sol than Sol 6.2 - which would explain why it's not as much better as we would've hoped.

It may also explain why their release is not going as well as they hoped. It's hard to predict how your first looped models are going to fare.

If you look at how long it took Google to recover from their failed 3.5 Pro, it would be surprising if GPT recovered from their rumored failed run on Astra in under a month. That would give them reason to release a looped Sol and call it Astra.

iamdelirium 3d ago on HN
Looped transformers have been known for years now. I don't think it's something new.
onlyrealcuzzo 3d ago on HN
They are JUST arriving to frontier models.

Even in the open weight Chinese models as small as DeepSeek V4.1 Flash - nothing is looped.

refulgentis 3d ago on HN
I think they're responding to the claim Astra can't be looped, there's clear and convincing research, as well as public comms, that it is. It's unclear why Astra can't be looped because looping is new, but, 6.1 Sol released a week later can be.
refulgentis 3d ago on HN
Astra was the first shipping looper. Unconvincing evidence would be reporting x OpenAI's quick reaction to it via a blog post from their chief research (it broke what's considered a safety taboo). Convincing evidence would be research from ex. Neel Nanda on serial reasoning, recent update at https://x.com/EpochAIResearch/status/2122762411287748696?a=J...
onlyrealcuzzo 3d ago on HN
Fable & Opus, Astra & Sol, and Carbon are all thought to be >1T param models - AFAIK, most people think Opus & Sol are ~5T param models and Astra, Fable are 10T+.

I think the rumor-mill consensus is that Google's models are significantly smaller while attempting to swing at the same level - but the ratios AFAIK are thought to be around the same - i.e. Neon is somewhere around half as big as Carbon, big enough that a newer checkpoint can make it temporarily a little bit better.

andai 3d ago on HN
The Bitter Lesson - Richard Sutton (2019)

https://news.ycombinator.com/item?id=19393432

weitendorf 3d ago on HN
There's only a limited amount of hardware for training and serving models at scale, and which form an upper bound for what model size/inference can support.

A GB300 rack has 72x288GB=20TB VRAM. Do the math and 10T parameter models are just about the max size you can easily train and serve (at 1M context) without doing a lot of custom hardware development or limiting your scale.

cl-coder 2d ago on HN
> Gemini 4 <Element>-lite

Surprised if it’s not going to be 4 electron/proton.

WarmWash 3d ago on HN
They are almost certainly being ambiguous so that people rely on the model descriptors to pick which one to use.

I think there are a ton of people who automatically default to the largest model with the highest thinking for everything. Which historically made sense but not nearly as much anymore.

At this point a model like hydrogen can cover just about everything an average person is doing. You don't need Carbon deploying an agent swarm pushing terabytes of parameters through HBM memory to help you plan your birthday party.

jeffbee 3d ago on HN
Wouldn't such users immediately run into quota walls?
WarmWash 3d ago on HN
They do, and then they complain.

Worse, people burning dumb prompts on large models also has the net effect of shrinking quotas for everyone.

DonsDiscountGas 3d ago on HN
I assume atomic weight. Neon seems to be better than Carbon, also faster and cheaper (all judging by the link)
ehsankia 3d ago on HN
Great, consumers now need to first learn chemistry to be able to figure out which model to use!
accrual 3d ago on HN
That's my thought too. If one needs a lookup table to understand the product by name then it's a product issue, not a consumer issue, even if everyone "should" have the first N elements memorized by atomic $property.
daemonologist 3d ago on HN
What about Argon though?
breakwaterlabs 3d ago on HN
Apparently even Google wants to pretend it doesn't exist.
Tadpole9181 3d ago on HN
But Argon has a higher atomic weight than both, and sits below Carbon and above Hydrogen. Also, Carbon is rated at better at long-horizon tasks while Neon seems to be smarter at contained work. So they're not strictly better/worse.
dudeinjapan 3d ago on HN
Neon is more noble.
adithyareddy 3d ago on HN
(S) Gemini 4 Hydrogen - fastest - $1.50/$7.50

(M) Gemini 4 Neon - second fastest - $3/$15

(L) Gemini 4 Argon - second slowest - $2/$10 (till Nov 30), $4/$20 (> Nov 30)

(XL) Gemini 4 Carbon - slowest - $2/10 (till Nov 30), $4/$20 (> Nov 30)

WarmWash 3d ago on HN
Argon and Carbon are the same class model. Carbon is just a newer rev of it.

There is a Helix model too (in the pipeline) which is the actual small size.

neop1x 45h ago on HN
Thanks. It's sad that you had to do the research and write this comparison here in a deep-buried comment instead of google having a single, clear landing page and proper meaningful names. Google totally messed up this naming.
whimsicalism 3d ago on HN
meant to be an opus/sol competitor, looks like. not fable/astra
Mogzol 3d ago on HN
As far as I can tell it's arbitrary and Google is just choosing element names they think sound good.

When Argon first came out it seemed like they could be going for noble gasses, like Helium (small), Neon (mid), Argon (big).

But then Carbon came out and it seemed like maybe they were going through the alphabet for model generations. Argon, Boron (internal), Carbon.

And now they've come out with Hydrogen and Neon, and there's no logical way to order them. The actual model sizes are Hydrogen (small), Neon (mid), Carbon (big). This doesn't match sorting by atomic number or alphabetically.

baoooooooooooo 3d ago on HN
Everyone seems quite upset/perplexed by the Gemini naming scheme but I quite like it. And hey, if you find yourself lost, the periodic table is probably the most ubiquitous ordered list in the world and serves to clear things up pretty fast. This will likely not be strictly true as models release over time, but is a good rough indication of what the team thinks we should expect of the models, in the same way the other labs run their naming.

I also like the personalities implied by these, with Anthropic saying “human literary achievement is neat”, OpenAI saying “space is neat” and Google being like “science is neat”.

Edit: I was coming back to say “I wish we could talk about stuff more substantive than the naming scheme” but (1) my comment may be objectively incorrect, and (2) I have to apologize that I haven’t helped the situation by… commenting on the naming scheme.

WarmWash 3d ago on HN
They are not following the periodic table in any discernible way though.
dudeinhawaii 3d ago on HN
They're in alphabetical order, almost like Ubuntu release names. There was a leak about a Gemini Oxygen model, which would further support alphabetical order -- skipping names they don't want to use.

So, names don't align with capability but instead "release freshness". This is unlike every other provider but it is what Google does (did?) with Android releases I believe (dessert names).

WarmWash 3d ago on HN
They went form A to N in 4 releases. Initially maybe they planned to follow the alphabet, but I think that went out the window
Chilko 2d ago on HN
I suspect the missing letters are internal models or checkpoints that were not released (e.g. Barium), although the jump from Carbon to Hydrogen and Neon is quite large
wat10000 3d ago on HN
The periodic table isn't really helping. Hydrogen is the lightest, that much makes sense. Beyond that it doesn't seem to work. The elemental ordering is H C Ne Ar. The cost ordering is H Ne Ar C (with Ar and C being the same per token but I gather that C thinks more and thus costs more). The capability ordering seems to be H Ar C Ne. Argon is the highest on the table but is neither the most expensive nor the smartest.
georgemcbay 3d ago on HN
> And hey, if you find yourself lost, the periodic table is probably the most ubiquitous ordered list in the world and serves to clear things up pretty fast.

But the way google ranks the models by capability doesn't follow periodic table order.

Hydrogen - 1

Carbon - 6

Neon - 10

Argon - 18

It could very well be based on their density (g/cm^3 in standard conditions):

Hydrogen - 0.0000899

Neon - 0.000900

Argon - 0.001784

Carbon - 2.267

This would make more sense than assuming it is based on their periodic table order as it more accurately maps to how Google markets their capabilities, but is too esoteric to really be useful for clearing up confusion for normal users, IMO.

bertmuir 3d ago on HN
I think it's more vibes-based. The different periodic groups have different meaning.

Neon and Argon are descending down the noble gases, I'd expect their successors are Krypton, Xenon, Radon.

They skipped Helium and went straight to Neon, which makes me think they've reserved Helium as a lightweight element, and are going to follow the lightest elements for the cheapest models. Hydrogen, Helium, Lithium, and so on.

Carbon reads more as a "foundation of life" kind of metaphorical choice, and they have some choices in the periodic groups nearby to keep that sort of metaphor for a while. My guess is they move horizontally - Nitrogen, Oxygen - then on to Silicon when they hit something appropriately transformative with the model.

I'd expect the most experimental or fastest models to eventually release on highly reactive groups - Fluorine, Sodium, etc. - and that they're leaving the metallic elements in the middle for differentiation in industry-specific fine-tuned models.

pavo-etc 3d ago on HN
Lmao no. The most ubiquitous ordered list is... numbers. Or the alphabet.
kphorn 3d ago on HN
They use the very precise "it sounds cool". Neon is 75% the price of Carbon but has 4 extra protons in the nucleus and is also a noble gas. Argon is a heavier noble gas at 16 protons.

element/model - atomic number - cost H - 1 - $1.50 & 7.50 per mtok C - 6 - $4 & 20 per mtok Ne - 10 - $3 & 15 per mtok Ar - 16 - $2 & 10 per mtok

Claude - length of story but fable messes it up Haiku - 3 lines Sonnet - 14 lines Opus - a novel? Fable - 30 lines?? Mythos 15700 lines?

Luna - 28 days Sol - 365 days Astra - 5-6000 stars

daemonologist 3d ago on HN
I have to assume they're messing with something other than size here - nobody would be this bad at naming different model sizes. Maybe sequence length or post-training strategy or sparseness or something else with the architecture.

(Although why they'd release all ~four is a mystery. I guess they're having a hard time picking the "best" one and so are just putting them all out to let the public decide.)

scotty79 3d ago on HN
You can probably infer that form tokens per second.
ariwilson 3d ago on HN
so which model am I supposed to use for what? someone help me

and why can't I use any of these with agy and a $20 plan? unlike openai and anthropic

pilooch 3d ago on HN
So this is above carbon ?
maplet 3d ago on HN
> Independent evaluations rank Neon as one of the top models and on benchmarks like the Vals Index, it ranks #2 (70.9%), and reaches #1 on both the Legal Research Bench and Tax Agent Bench.

On Artificial Analysis Intelligence, it scores comparably to GPT-6.2 Sol but is significantly more expensive.

onlyrealcuzzo 3d ago on HN
This model feels like it might actually be as good as Opus 5.5.

This is probably the set of benchmarks most people care about: https://www.vals.ai/comparison?m=claude-fable-5-2,cobalt-c-h...

Overall, performs just under Fable (above Astra or Sol 6.2 or Carbon - slightly ahead of Opus 5.5).

Latency is by far the best for the level of quality.

It's about 2x more expensive than Sol 6.2, but significantly better and faster. You get what you pay for.

catigula 3d ago on HN
0% chance this is better than any frontier OpenAI model.
HDThoreaun 3d ago on HN
How can you be so confident about a model that was released today? I just dont get the lack of epistemic humility here
catigula 3d ago on HN
>how can one presume the sun will come up tomorrow merely because it has come up every day prior? have some epistemic humility! i am a very smart rationalist.

Edit: I tested the model and it found several mistakes Astra 6.2 made. It’s possible…

WarmWash 3d ago on HN
I think that OAI likely has the strongest internal model, but I don't think they can figure out how to ensure alignment on it
f13f1f1f1 3d ago on HN
I also have the hottest wife but no one can see her because people wouldn't be able to constrain their lust. I need to hide her to maintain safety.
WarmWash 3d ago on HN
However hot she is, she isn't solving millennium problems.
SuperV1234 3d ago on HN
I am a Claude Ultra subscriber that's very happy with Opus 5.5.

If I wanted to realistically try out Neon in some of my workflows without committing to a paid subscription before seeing if it can actually work as well as Opus, do I have any option, or do I just need to take the gamble?

onlyrealcuzzo 3d ago on HN
They say it's going to be available for Pro users - so you don't need a $200 account (Ultra). But, I don't think you're going to be able to try it unless you get at least a $20 account.

IIUC, you also don't get to try Opus unless you have a subscription, right?

deaux 3d ago on HN
Reminder that Google trains on your data if you're using a consumer sub whereas Anthropic and OpenAI claim they don't if you've got all relevant boxes turned off. I know everyone stopped caring about this at the start of the year and believes the latter two don't honor it anyway, but it's good to know.
user43928 3d ago on HN
I was thinking of giving Gemini 4 Neon a try as a code reviewer instead of GPT-6.2 Sol, and found this in their FAQ:

"Using third-party software, tools, or services to access Antigravity is a violation of our Terms of Service"

and also: https://antigravity.google/docs/faq?can-i-use-antigravity-cl...

I'd like to keep my Gmail account, so Google can keep their model and will not get my business.

what 3d ago on HN
You don’t need to use the models through antigravity. You can get them through ai studio (or whatever Google calls it) and pay api pricing to use it via whatever harness you want.
WarmWash 3d ago on HN
You can't use subsidized tokens in 3rd party apps. Although they just suspend your Gemini account now, not your google account.

That being said, you can do whatever you want with API tokens.

user43928 3d ago on HN
You can with Anthropic and OpenAI, using 'claude -p' or 'codex exec'.

Paying for Gemini at API rate is therefore out of the question for me.

joshuamorton 3d ago on HN
That violates the ToS for both, and they absolutely will ban you (there's tons of examples of people complaining about their cool tool to get Claude Code shimmed over an API not working/getting them banned, same with OAI).
user43928 2d ago on HN
They won't ban you for that.

I never had any issue using Claude for code review in Codex and vice versa.

I remember that half a year ago, Anthropic planned a change to make claude -p draw from a separate usage pool at API pricing, but after feedback that change was scrapped.

What people were banned for is using the subscription authentication in a third party tool, using many accounts, or reselling.

Invoking the official CLI from another agent for legitimate personal use during development should not be a problem.

p_l 3d ago on HN
you can use agy in batch and/or JSON mode - it's what I do for few of my projects. You can also use "agy acp" to get ACP server now (used to be a separate binary)

The requirement is that you use antigravity harness to access subscription-based accounts, not that you can't wrap it

user43928 2d ago on HN
Their official wording seems to be that it's not supported:

> Wrapping agy inside third-party coding agents or applications, or automating account rotation, is not supported under consumer plans.

At least they no longer ban the whole Google account, so I think interpreting that as a prohibition only for abuse would be reasonable.

fastest963 3d ago on HN
I think that page is out of date. You can use it via T3 Code now and Google has said it's okay. It's via their official ACP-compatible endpoint that's now provided by the antigravity-cli.
koolala 3d ago on HN
You can install Codex in Antigravity its kinda fun running both.
what 3d ago on HN
You can always pay API pricing if you don’t want a subscription?
desterothx 3d ago on HN
openrouter or api directly? great question to use your claude sub to figure out easily :)
layoric 3d ago on HN
How do you find Fable 5.2? I tried it for a few days for coding, and I'm back on Opus 5.5, and even Deepseek 4.1 Flash for simple stuff. It was just going off on tangents way more for my workflow.
krisroadruck 2d ago on HN
Glad I'm not the only one. I found Fable 5.2 to be a substantial step back vs Opus 5.5. It would happily burn a weeks worth of Ultra usage in a day to output an absolute slop pile of wrong assumptions and bad decisions. It just sucks for agentic use or even review work. Initial design back and forth is the only place I could get it to be useful. I'm starting to wonder if all the amazing examples anthropic gives for it only work if it has effectively an unlimited token budget so it can spin it's wheels for however long it takes to eventually get a decent result.
DeathArrow 3d ago on HN
You can use Neon on OpenRouter.
anentropic 2d ago on HN
OpenRouter
dudus 2d ago on HN
You want a Google AI Studio API key.
lanthissa 3d ago on HN
looks good gemini team fixed the massive spread they had vs openai and anthropic in terminal bench - swe bench.

That spread is my personal bs detector for models over promising.

andriy_koval 3d ago on HN
I suspect all labs just benchmaxed those benches by now.
InsideOutSanta 3d ago on HN
Are these names a practical joke Google is playing on us, or are they actually serious?
maplet 3d ago on HN
My guess is that Google thinks the trade-offs are too nuanced to have a simple small/medium/large ordering. Neon is between Hydrogen and Carbon in terms of speed and cost, but it scores highest on Artificial Analysis Index and has a xhigh thinking level (unlike Carbon).
kristjansson 3d ago on HN
So now we've got to get a periodic table out to understand Gemini releases? Hydrogen is the weakest, and Argon is the middle, but Carbon is super-Argon and Neon is better than both? And I can't even look at the price list if my logged-in google account doesn't have AI Studio(??) access?

e: No, Neon is the cheaper, mid-tier between Hydrogen and Argon. So Argon/Neon are aligned along the noble gasses, and Hydrogen kinda makes sense as the weakest ... but they threw Carbon in without any consideration

Melatonic 3d ago on HN
Maybe Carbon is actually just two Pelicans in a trenchcoat
numpad0 3d ago on HN
Hydrogen < Neon < Carbon? But where's Argon? Why those elements?

  01 02 ... 13 14 15 16 17 18 
  H                        He
  Li Be ...  B  C  N  O  F Ne
  Na Mg ... Al Si  P  S Cl Ar
???
ajs1998 3d ago on HN
There is no logic. Someone made a list of cool element names and when they run out they'll switch to some other scheme that communicates nothing
thehappypm 3d ago on HN
It’s alphabetical chronologically
deaux 3d ago on HN
The fact that there's a dozen comments on this post alone talking about periodic tables, element sizes and other stuff and these are the top comments and you and me are seemingly the only ones noticing this pretty obvious pattern..

It's not like this has no precedence. Google themselves are the best known example in software with Android codenames. Starting with [A]rgon was a clear signal in and of itself. But above all, just basic common sense strongly goes against Google uses some very limiting, niche pattern rather than just alphabetic.

I'd chalk it up to eternal September but I'm not convinced most of these people are new. A particular brand of neurodiversity, I guess.

Traster 3d ago on HN
It begs the question, if they've released Argon, Carbon, Hydrogen... did they not release, Boron, Deuterium through to Gallium, etc. Were they using Lithium internally and decided they couldn't release it for PR reasons? Or did they just think we're too think and will get scared if they start releasing scary sounding elements.
numpad0 2d ago on HN
Sorting elements alphabetically! How would you handle edge cases like Natrium? Maybe I'm just dense[1], but it feels sorting a-Z might be an American behavior.

1: is this use of "dense" obsolete given dense LLMs are often smarter than sparse models?

wernam 3d ago on HN
Maybe Knight moves on the periodic table :)
jwpapi 3d ago on HN
LOL
seizethecheese 3d ago on HN
I got tripped up assuming bigger was a larger element. Seems like it's vibe, like Carbon feels bigger than Neon.
clutter55561 3d ago on HN
Surely the AI companies are colluding... to make sure no one can make an informed about model class/size/effort.
xnx 3d ago on HN
It won't be good for Anthropic's stock price if investors realize it's always been a three-way race and Google is on a trajectory to overtake.
hn8726 3d ago on HN
I'm very excited for any model that performs on par with Opus 5.5 but doesn't speak claudish. It increasingly looks like Google might get there soon
Ternari 3d ago on HN
They’re already there. Have you not used Carbon or Neon yet?
agrippanux 3d ago on HN
All I really want now from any provider is Opus 5.5-level at blistering speed and a fraction of the cost. How close does this come?
moecables 3d ago on HN
It's crap how Google doesn't show pricing on the pricing page for Antigravity: https://antigravity.google/pricing
raincole 3d ago on HN
It looks like argon is deprecated.
doodlesdev 3d ago on HN
Already removed from the Gemini app. Replaced by Neon.
vintermann 3d ago on HN
That was quick. Really hope this new one hasn't regressed on historical HTR.
bemtevi 3d ago on HN
If one day, AI becomes provably good at all programming tasks, at working at mathematics better than all mathematicians, and at solving problems in almost all domains with maximal efficiency, I think it'd be a dystopia.

What I simply don't get is why people want to push the boundaries of efficiency and knowledge. I used to think as a teenager that all knowledge was good but how good is knowledge when we can't acquire wisdom with a comparable speed?

The greatest thing about the early internet and earlier science was that people shared and communicated with each other. Now they just share what they produce wihouth much humanity at all.

It seems to me that the creation of all these models has just become a race where no one really cares about the finish line and no one really understands why we're doing it, except for the short-term trail of dollar bills on the race course.

For me, each new model release is a sad day for humanity.

simpsond 3d ago on HN
We are already there, and yes, it often feels inhumane.
numlock86 3d ago on HN
> If one day, AI becomes provably good at all programming tasks, at working at mathematics better than all mathematicians, and at solving problems in almost all domains with maximal efficiency, I think it'd be a dystopia.

Which of these are missing in your opinion? I'd argue we are there already. At least very very close, as "with maximal efficiency" is quite an arbitrary term.

bemtevi 2d ago on HN
True. We're getting close, though AI-assisted programming still needs human direction for complex projects and AI-assisted mathematics has trouble creating totally novel ideas. But yes, it's close and a miserable situation to be in.
bertmuir 3d ago on HN
Because hundreds of millions of people live in abject poverty, and if we divvied our economic production up evenly today there'd only be ~$5,000 to ~$15,000 per capita at a US cost of living i.e. the global economy is far too small for everyone to have good lives.

The growth mindset isn't really for you, if you're already rich, and don't want to be richer. It's coming from the vast majority of the world, which is poor, and the minority that is rich but wants to get richer.

esperent 2d ago on HN
> at a US cost of living

We don't want a US cost of living, we want something like a Western European quality of living, and we want to figure out how to get that across the world even for places that have much higher populations than Western Europe, and at a low environmental impact.

Comparing how much a decent quality of life currently costs in the most expensive countries isn't useful.

Also, we need to focus on the basics rather than just trying to give people tons of money: good food, decent housing, free healthcare and free education.

bertmuir 47h ago on HN
I think most people here know what the price levels in the USA is like, so it's clearest for communicating what our economy is currently able to produce.

Do you think "~$4,000 to ~$12,000 at a German cost of living" would have made it clearer how much we actually have to go around today? Or maybe just saying "the global economy isn't even close to the size where we produce enough for everyone to have decent food, housing, healthcare and education"? To me that seems likely to interpreted as vague and easily dismissed.

novaRom 2d ago on HN
Don't worry - AI regulations are coming soon.

A major point will be complete training-data transparency with independent expert review. In EU a statutory ban on collecting/storing/using users' data for training (including the "de-identified data" loophole), except voluntary explicit opt-in.

I wonder if a mandatory disclosure of training recipes paragraph will be included, likely in EU only (e.g. with tiered-access design to counter trade-secret objections) ?

dprkh 3d ago on HN
Has anyone else been having an issue where the models would answer its own question? I have been seeing this quite a lot with Gemini 4 models.
jeffbee 3d ago on HN
Oh yeah, and you're paying for it to say "wait. what about [[thing]]?" then answer itself.
smallerize 3d ago on HN
The "long-horizon" ones are optimized for this. They won't get a high score on the benchmarks if they stop to ask questions.
dprkh 3d ago on HN
This seems to be specific to Gemini because I have not experienced this with OpenAI's models. In fact, Codex now has a feature such that the model can ask a question and continue making progress on work that is not blocked while waiting on your answer.
smallerize 3d ago on HN
I wasn't benchmarking or anything rigorous, but I noticed it with GLM 5.3 Flash. It was downloading and installing stuff to my computer to try to work around an assumption in the prompt that it knew was wrong. I also noticed it trying to work around websites' bot prevention measures. I stopped it often.
vintermann 3d ago on HN
And I hate it.
Jorge1o1 3d ago on HN
This is some of the most confusing naming I've ever seen. There's now: Gemini 4 Hydrogen, Carbon, Argon, and Neon!

Maybe the atomic number correlates to model pricing? Nope! Neon (10) is marketed as cheaper than Carbon (14).

Maybe the atomic number correlates to overall intelligence? Nope! Neon (10) and Carbon (14) are both marketed as smarter than Argon (18).

The only thing I can think of right now is that it's in alphabetical order by release... but in that case why the periodic table?

dannyw 3d ago on HN
Major props to Google for announcing a model and releasing it on the same day.

Unfortunately, their recent change to allow me to opt out of AI training (which would also include connector data like Gmail, photos, etc) hasn’t rolled out to me yet[1]; so I won’t be upgrading my subscription; nor using the Gemini app until I can opt out.

[1]: https://arstechnica.com/ai/2026/11/google-will-let-you-opt-o...

testtestmetest 3d ago on HN
Oh know, they're going to steal the code for you 3 MAU SaaS !!!

/s not trying to be mean, its just weird hill to die on. Every company is stealing all your data in ways you unknowingly consent to at every second of the day.

5555watch 3d ago on HN
"Gemini History" seems to only target Gemini App functionality, what about Antigravity chats?
nixosbestos 3d ago on HN
well, there's https://antigravity.google.com (not to be confused with https://antigravity.google)

Which ... doesn't necessarily show what you want either. But you can access "remote history" in bottom left. Maybe, if it works for you.

sdvs9so 3d ago on HN
I think that just changed as of a Nov 16th. You're now able to opt-out your chats from model training while retaining chat history[1].

[1]: https://blog.google/innovation-and-ai/products/gemini-app/up...

deaux 3d ago on HN
> Currently, if you disable history, Google still retains that data for 72 hours (and 24 hours for the user). Once the updated page is live, this duration will change to two months.

So a regression then.

abraxas 3d ago on HN
Is it true that this one has a 1MM token OUTPUT window unlike the other competing models that are capped at 128K?
Computer0 3d ago on HN
With agents often taking many turns before responding to you, what use case does this change for you?
abraxas 3d ago on HN
I think for large refactoring jobs this will open up a new category of work that an agent will be able to do. Also reasoning steps count towards output tokens, no?
Computer0 3d ago on HN
That is a good point, I hear of sonnet frequently hitting a limit in its thinking.
xiaoyu2006 3d ago on HN
When you think the naming scheme is already bad enough... It can always get worse.
artdigital 3d ago on HN
Okay so it seems like the ranking here is Hydrogen -> Neon -> Carbon even though that is not clear at all from this release. At least Googles model picker in the Gemini app now lists those 3 above each other in this order, and keeps Carbon for “Code & complex work”

So Neon is the new default, and Carbon the slow deep thinking “when you need it”? A bit like how Opus 5.5 was better than Fable 5.1 in most things, but Fable still had the edge over deeper things (then 5.2 came along)

Hydrogen has already became my default for 80% of small tasks. If Neon is as fast but smarter that would honestly be very cool

gradus_ad 3d ago on HN
To be fair prior to Gemini 4, the app had 3.1 as the "Complex work" model while 3.8 Flash, far superior in every way to 3.1, was listed as the "Fast" model.

So I'd take their hierarchy with a grain of Sodium.

vtail 3d ago on HN
Do you know what happened to Argon? Was it a flop?
shric 3d ago on HN
For some reason nobody reacted to it
lukax 3d ago on HN
Carbon is Argon with extended thinking
artdigital 3d ago on HN
It was good, but Carbon quickly superseded it
Robdel12 3d ago on HN
Their pricing page has no.. pricing? Am I the only one that sees that?
bisonbear 3d ago on HN
Anthropic competing with Google instead of OpenAI at the frontier is unexpected, I have been doubtful of Google DeepMind's status as a frontier lab in the past few months, but damn they've been cooking recently
Tadpole9181 3d ago on HN
It's important to note that DeepMind and the other two labs aren't playing the same game.

The latter need to raise funds and IPO, but the former is an arm of one of the most powerful companies to ever exist with some of the largest data warehouses around and custom hardware. And they have an endless pile of internal code to maintain already with hundreds of millions of end users already bought-in.

The incentives being different leads to different focuses and product lifecycles.

serioussecurity 3d ago on HN
Billions of end users.
polski-g 2d ago on HN
Not unexpected at all. Google already had the most popular LLM model in the world, they already had the datacenters/power, they already had the data in Google Cache. They just needed to set up a training run for a bigger model.
45sdasf45 3d ago on HN
to paraphrase xkcd 2501: "Its easy to forget the average person only knows the atomic numbers of the noble gases and one or two nonmetallic elements".

Say what you like about Apples Pro, Max and Pro Max -- at least it gives some indication of size to the average person.

bluegatty 3d ago on HN
There are too many names and versions to keep up with.

They all need to pick a dam naming scheme and stay with it.

gradus_ad 3d ago on HN
So Anthropic is being valued at around $1.2T and their only revenue stream is a collection of frontier AI models with 0 moat that are basically as good as Google's, a company which has the luxury of other revenue "streams" (rivers/seas/oceans) and can certainly win any price war with Anthropic.

Can't wait to buy Puts on ANT...

impulser_ 3d ago on HN
Is there any theory on what is Anthropic's secret sauce? I have been using the latest Gemini models and they just don't seem as good as Opus 5.5 despite the benchmarks. Yeah it comparable on some tasks maybe slightly better but as an overall general intelligent model nothing compared to Opus 5.5 and it hard for me to understand why. Opus feels like human level intelligent like I'm talking to a coworker basically while GPT and Gemini feel like I'm talking to an AI and have to change how I talk. It hard to explain and maybe I'm the only one that feels this.

Is it because Anthropic put a ton more effort into the character of the model that it basically roleplays as a human and that why it tends to think better when giving tasks that it might not be trained heavily on.

I have used Carbon more than most model to try and get the same results but it just feels like i'm talking to a smart AI. Which is fine if you are asking questions, researching an idea, but if you actually want to work on something, build something, it just falls apart. Like Opus can run for days on a task and it feels like it understand what it was doing 2 days ago even when it might not be in it context while other models feel like they breakdown and get lost or falter as the task grows.

baq 3d ago on HN
Opus 5.5 leads the CanTellForestFromTheTreesBench 42.0.
jorl17 3d ago on HN
Here's my average experience with Carbon.

(After a long conversation with existing messages)

User Prompt: Ok, so SSH to the unraid server and update the plex db

Reply:

I need to think how to answer this.

The user is asking me to SSH into the unraid server. This seems to imply there is a machine to which we can SSH to. I should look for places where hosts may exist.

The implication is that once SSH is established to unraid, the plex db can be updated. It seems plex is a service that the machine has in which we can run updates.

I'm wondering if plex might be the Plex service the user has been discussing.

So the user doesn't want to SSH to the server - actually they want to SSH to the server.

... Formulating a plan ...

[some messages later]

Aha! THIS IS THE SMOKING GUN! `~/.ssh/config` HAS THE REQUESTED HOST!!! IT'S TIME TO CONNECT!!

[some messages later]

Ok, time to act. I'll ssh to the server and follow coherent rules.

----

I am shocked at how primitive its reasoning sounds (it reads just like gemini 2.5-pro's old reasoning). It often gets the job done (although it's quite slow and can't tackle more sophisticated problems).

Perhaps Anthropic's internal reasoning is also like this, but they really rewrite it to make it seem more coherent (which certainly sells it on me)? I almost want to click stop just by reading this nonsense.

aenis 3d ago on HN
One of my long horizon tasks benchmarks involves writing logic controlling a team of soldiers in a simulated environment. Argon and Carbon both made such discoveries such as "it would seem that i need to be pointing at an enemy to shoot them". Its almost entertaining to read, its like its discovering the world on first principles.

(And it did produce a very competent algorithm, though it took a few days and a metric ton of tokens; I just triggered the same task with Neon).

jorl17 3d ago on HN
I'm curious about how it goes with Neon! This night I'll leave Neon with some lofty goals on a project and I'll see how it fares too.
aatd86 3d ago on HN
lol Aenis the god.
iamdelirium 3d ago on HN
All displayed reasoning is not real reasoning for all models now. It's fake since they can be trained on.
jorl17 3d ago on HN
I want to believe you, and it's what makes the most sense! But if that's the case, why is google (intentionally?) presenting such terrible reasoning traces? To make sure that if they're trained on, they're trained on garbage? I guess it makes sense, but still...

I was not really fabricating the example I gave too much. Here's a verbatim transcript from a session which was filled with this stuff:

    ... thought LET'S READ RecordReader.cs RIGHT NOW!! RecordReader.cs IS WHERE EVERY SINGLE TABLE, ReadPlan, DecideValueVisibility, ReadRowCounts , AND ShowValues READS ITS DATA!! Let's read RecordReader.cs (lines 1-300)!
8note 3d ago on HN
reasoning traces also arent real reaaoning. its what tokens warm up the outpt the best
dist-epoch 3d ago on HN
What you're seeing is a small model summarizing and possibly obfuscating the real thinking.
jorl17 3d ago on HN
Yes, I understand it.

It seems rather ironic that Anthropic givens the most logical reasoning traces, though. You'd think if they're ahead they'd either hide them or make them produce garbage (like what seems to come out of Google's models).

orbital-decay 3d ago on HN
Neither has any relation to the actual reasoning trace. The summary is actively misleading and you'd be surprised how different it is from the raw CoT of the model. Just ignore it.

If you want to see example traces without jailbreaking claude or gpt, see open models that display the CoT without trying to hide it. It's pretty useful sometimes, with usual caveats (it deceptively looks like English but is heavily reward hacked, and it's generally a weak proxy for model's reasoning process).

jorl17 3d ago on HN
This makes sense. Nonetheless, I find that reading Anthropic's reasoning traces is sometimes a trigger for researching more stuff. Often I'll see models thinking (or fake thinking, whatever) about something and I'll be curious to check out more about that topic if I'm not particularly familiar with it.
jrowen 3d ago on HN
FWIW, you're not wrong. As I said in my other comment, I don't really know what other models are up to but Claude is definitely demonstrating a deep level of understanding, and I learn things from it all the time.

Sometimes I'm like...is everybody else using the same thing that I am? This is insane.

They definitely let it churn until it figures something out, it's not uncommon for it to take 15-20 minutes to answer, is that normal? They just have "relentlessness" maxed out and some kind of second-order structure to orchestrate a bunch of agents into one unit of work.

jorl17 3d ago on HN
That's the exact same feeling I have.

It's obvious that what we're getting are summaries. Sometimes I think I even see vestiges of the summarizing model instructions. But, that being the case, it's bizarre why Gemini's summarizer acts like a toddler, and Claude's actually lets me learn and, more importantly, stop it at its tracks when I see it's going down the wrong path.

On the other end of the spectrum, if you toy around with the open-weights models, you see the insanity that is going on in their output. I can totally believe in a scenario where claude is basically doing that and then summarizing it into something coherent, getting rid of the 10 opposing viewpoints, "aha!" and "Hmmm, but wait!", but it does it elegantly.

I've tried constantly to use other models (and I do), but Claude spoils me to no end. I keep thinking I must be holding everything wrong.

disgruntledphd2 2d ago on HN
> On the other end of the spectrum, if you toy around with the open-weights models, you see the insanity that is going on in their output. I can totally believe in a scenario where claude is basically doing that and then summarizing it into something coherent, getting rid of the 10 opposing viewpoints, "aha!" and "Hmmm, but wait!", but it does it elegantly.

Yeah, the open weight models (the only ones that show the reasoning tokens) are just absurd. Wait but why oh wait now I have it, this makes total sense!

I find it kindof endearing, tbh as I definitely have similar trains of thought when developing.

jrowen 2d ago on HN
> summarizing it into something coherent

This is like, the entire ballgame. This is what effective communication is, this is what understanding is. Anyone who can do this, man or machine, would do so, I would think.

deaux 3d ago on HN
Off-topic, but this kind of conversation tells me just how little the average HNer still uses LLMs outside of specific coding tasks, or at least pays attention to them. Which in turn explains so many comments. Probably a good thing though.

You're of course completely right. Interestingly the "summaries" have only become more diverged over time. I now believe the big 3 are barely summarizing at all any more. I deem it likely that the small models don't get the CoT at all, they just completely make up something "plausible" solely based on the user message with a little bit of previous context. Much of the time it will be the complete opposite of something I asked for, or barely related.

Anthropic has changed this up multiple times over the last 12m, I'm talking about the iteration that was around the longest this year. I think they changed it again within the last feww weeks and potentially started actually summarizing again.

Your suggestion is the right one. Everyone who uses LLMs for their job should spend some time with open models like Deepseek and see what the thinking actually looks like.

marmarama 3d ago on HN
> Everyone who uses LLMs for their job should spend some time with open models like Deepseek and see what the thinking actually looks like.

I'm not sure DeepSeek 4.x is a good example; its reasoning traces are extremely verbose and detailed. Which is probably one of the reasons it performs so well, and appears to be a deliberate architectural decision to improve skill at the cost of token usage and time-to-answer.

A better comparison might be Gemma or Qwen. Running these locally and with no summarisation, the reasoning traces are not all that different to the big proprietary models in terms of verbosity and content, even if they're not as smart.

deaux 2d ago on HN
> I'm not sure DeepSeek 4.x is a good example; its reasoning traces are extremely verbose and detailed.

That's why I recommended it. It shows that for a given model, it's CoT might look different from what people expect. I also named it because I saw in another comment here - or maybe it was on a thread about Carbon, one of the two - chatter about Gemini 4 outputting reasoning inside its user-facing message and looking like "I'm not sure. Maybe it's A? YES! Now I get it! ...", which is of course the style that Deepseek 4.x's traces also have.

jrowen 3d ago on HN
I mean, it does though. The sequence of thoughts that Claude goes through in response to my queries blew me away the first time I used it, and it continues to impress. Just the act of putting those sentences together demonstrates such a deep understanding of what I'm saying. If it's just an approximation, whatever it's actually doing is even better.

I haven't even really tried to use other models in the same way, and this thread makes me feel like I don't need to.

behnamoh 3d ago on HN
Not only that, Anthropic injects fakes reasoning and tool calls in there to make it harder for distillers. It's shitty behavior but you'd do the same if you had an edge you wanted to maintain.
ehsankia 3d ago on HN
Every day I'm just more and more amazed at Opus 5.5 and how ahead it was. Even with the insane pace we're going at, it's amazing that two months later, we're barely putting out models that hit the price/speed/quality of Opus 5.5

No wonder they went for the 5.5 naming, instead of the .1 increase. From just the benchmarks, it seems like neon is basically on par with Opus 5.5 which came out 2 months ago.

aenis 3d ago on HN
Same here. Spin up a few long horizon tasks, have it manage the whole thing as a project, you can talk to it, interrupt it, engage in side-quests, its like talking to a human.

Have another session on another machine (say, IOS build on a mac, and server build on a linux?). No problem, they will happily coordinate with each other through the /remote-control protocol, and then you feel like you are in a crowd, with those sessions talking to each other and over you.

No such feel with OpenAI nor Google models for me. They are absolutely great for lots of things, but nothing comes close to the feel of Opus 5.5 and Fable.

(I am biased; i mainly use the anthropic models so i got used to them).

jwpapi 3d ago on HN
I dont even feel like fable 5.2 is better. Imho the best model rn is Opus 5.5 MEDIUM.
p1necone 3d ago on HN
The first time I tried to use fable 5.2 it made a bunch of really amateurish technical design assumptions about my project that cursory reading of docs would have disproven (but also should have just been discarded because of taste reasons anyway). I switched back to Opus 5.5 after that.
jwpapi 2d ago on HN
Yes there were even benchmarks by humanlayer how medium leads to better codebases and on multiple sessions on the name codebases (including reverts back and force etc)
ravikapoor101 3d ago on HN
I had the opposite experience.

I had subscription to both anthropic and openai. I was out of credits for openai, so used anthropic to fix some coding issues. Spent whole weekend prompting while it kept failing over and over. Finally on Monday I had openai's weekly quota start, 2 hours issue was done.

The same thing happened again, after a couple weeks. Two weekends wasted trying to make Anthropic models to behave.

aenis 2d ago on HN
I'd think its maybe a question of getting used to a given model, developing good intuitions on how to prompt, divide work and so on? I have spent probably 2000 waking hours this year with Anthropic models, and maybe 200 on openai, so i naturally consider them easy to work with. My own benchmark suite probably reflects the way I work, which is why its mostly Anthropic models getting the top scores etc.
ehsankia 3d ago on HN
fable 5.1 was strictly worse than opus 5.5 (which was embarasing). Even Fable 5.2 is slightly better, but both slower and more expensive. Again, pound for pound Opus 5.5 from 2 months ago is still probably one of the best models.
aenis 2d ago on HN
To me the revelation was how good Sonnet is. But yeah, my daily driver for anything is Opus 5.5 and I pray to the gods they won't nerf it anytime soon.
pjankiewicz 3d ago on HN
I agree Opus 5.5 is something different entirely. It is even hard to find tasks for other agents to be honest where you know the results will be picked apart when Opus 5.5 reviews them. In my setup Opus 5.5 uses codex from time to time when I'm running out of credits and it decided to run 2 astra agents for the same task to improve its performance.
behnamoh 2d ago on HN
> I agree Opus 5.5 is something different entirely.

If model versions are anything to go by, I'd have assumed Opus 5.5 and 5 were built on the same pretrained core, but the vastly superior performance of 5.5 makes me wonder if Anthropic literally pretrained a new one using Fable (following their RSI advances).

Opus 5.5 may be the first RSI'd model and it shows.

nebben64 36h ago on HN
I'm a noob; what do you mean by this, like how does your Opus agent know when to instantiate astra agents and how does it do it / communicate with it?
pjankiewicz 19h ago on HN
It does it through codex CLI. It is basically a tool that accepts a prompt. There is a way for claude to chat with codex as a person would do but thats slightly complex than a CLI tool.
nicoburns 3d ago on HN
> two months later

Mad that this is seen as a long time horizon!

ehsankia 3d ago on HN
I agree, but we live in an age where there's new models announced Weekly. We've had Argon, Carbon, Hydrogen and Neon all after Opus 5.5.
Snibz 2d ago on HN
Bit of a strange thing to be amazed by. I certainly agree that Opus 5.5 is a strong and useful model, it's been the model I've used the most for work in the past few weeks. But is it really amazing? Any other arbitrary AI company could equally come up with arbitrarily advanced models, applications, whatever. I'm certainly not amazed by the fact that Anthropic provides access to a strong frontier model.
jaccola 3d ago on HN
Claude annoys me with condescension sometimes... makes it feel far less like a 'tool' and can even come across as a bit judgemental!

Just today I was filling out a security questionnaire and it said "I won’t write a PDF that claims zero credentials while this token is in the logs." But we were in the process of removing this from the logs so it was perfectly accurate to write and I wish it had deferred to my explicit request...

I think Anthropic overcorrected on the sycophancy; Google models to me seem a nice balance of highlighting issues / risks while still doing what I actually ask.

enraged_camel 3d ago on HN
Claude models will go to extraordinary lengths to ensure sensitive data does not end up in their context window. It's a bit nuts. But frankly I appreciate it a lot. Even when it is annoying I know the model is doing the right thing.
impulser_ 3d ago on HN
Yeah, I notice this as well. I was building something to help me understand logs from the different coding agents and it wrote redaction code to redact secrets out of the logs just in case if the logs had any so it wouldn't see them.
hadlock 3d ago on HN
Opus 4.x really struggled with secrets rotation in sops, even in 5.5 it's constantly emitting secrets to the shell history. Thankfully everything gets tested in dev, but Claude models are the reason why we have such good secret rotation now.
aenis 3d ago on HN
Anthropic models won't lie. When we fix security issues identified with sth. like Wiz and are writing a report, it won't write a summary suggesting that something we just fixed wasn't broken (even if that thing never made it to production). It has something akin to moral standards :-)
willis936 3d ago on HN
I've seen them ignore clear directions that in documentation it's supposed to read after a compaction. That seems a lot like lying to me.
epolanski 3d ago on HN
> Is there any theory on what is Anthropic's secret sauce?

Lots of PR it seems?

I have all of Ant/OAI/OpencodeGo subs and test lots of models and I just cannot understand how and why some people find specific models so ahead, when it seems to me like reality is so nuanced and so specific to the particular task and how you approach it.

WarmWash 3d ago on HN
Never underestimate the power of minimalist hand drawn graphics on an off white background with a clean typeface.
deaux 3d ago on HN
More like, most people especially on here do 90% of their "tests" on very similar tasks - coding on specific codebases - and the other 10% is just asking stuff in chat. So they never get to experience the jagged edge. Anthropic models could be horrific at translation to Nahuatl or greenhouse cherry tomato management compared to Gemini and OpenAI but there's one person on HN doing such things and tens of thousands doing enterprise SaaS coding.
novaRom 2d ago on HN
They have tripled their PR team during just 2025 to roughly 80 people, today numbers can even be larger.

Btw, just one B200 x 8 server is all any mid-size company needs: any open source harness + any 80 - 120B active params model - it will fly and serve hundreds of agents. All data and processes are yours - no subscriptions, full control.

surround 3d ago on HN
I suspect that Anthropic has more user data to train on from Claude Code. Claude Code became popular sooner than Codex and Antigravity, and remains more popular - thus more data to collect.

Cursor was popular before any of those three, and just look at how much Grok improved after they acquired Cursor's data.

behnamoh 3d ago on HN
> Cursor was popular before any of those three, and just look at how much Grok improved after they acquired Cursor's data.

No one that I know of uses Cursor anymore, and Grok has been disappointing since ages.

surround 3d ago on HN
I don't disagree.
teknologist 2d ago on HN
Grok Bot seems relatively popular, and it does launch and use Cursor and Cursor Agents.
disgruntledphd2 2d ago on HN
> Grok improved after they acquired Cursor's data.

I never used Grok before the Cursor acquisition (we use Cursor mostly), and if this is improved, I shudder to think what it must have been like before the acquisition.

It's like an irritating Sonnet 3.5.

foohis 3d ago on HN
Anthropic spent a lot more time editing their pre-training data with higher-quality material. For example, they used actual books for pre-training, because books are generally run through human editors for higher quality. (this is from the copyright lawsuits)

Other companies would use completely raw internet data for pre-training, which is basically trash. Anthropic's secret sauce is that they filtered and edited down their data for only the highest-quality material. You really don't need quadrillions of tokens to pre-train, as that's just random noise. A smaller-but-higher-quality data set is better in every way.

Meanwhile its obvious that companies like Google would use trash data like email and chat logs or Internet forums for pre-training. Just complete trash data.

This is a classic case of "less is more" and "garbage-in-equals-garbage-out"

quixoticelixer- 3d ago on HN
That doesn't really make sense because as good as Opus 5.5 is it writes awful terrible, disgusting slop.
foohis 3d ago on HN
It's a much, much better writer than literally every other model.
quixoticelixer- 3d ago on HN
It's better than Opus 5 but it still has that incredibly punchy contrasty LLM style that worse than most other models
thriving_earthw 3d ago on HN
I sometimes use Gemini for quick searches or stuff and the writing style is such a breath of fresh air. I mean I've been so locked in to the anthropic ecosystem because I find that code, the desktop app, and the ability of their models to understand what I mean and execute so well is incredible. But sometimes I'll read something it writes and I'll want to throw my computer at the wall. I just become apoplectic with rage for a second before I of course cool down and move on.
BigGreenJorts 3d ago on HN
I've been feeling this more and more lately as I've been trying to delegate technical writing to it. It's so horrifically repetitive, verbose and somehow non specific I can't even really use it as a base to edit down to something legible.
copperx 3d ago on HN
It's sad that you can't quite delegate the writing task to another model, because Claudese is highly contagious.
plmpsu 2d ago on HN
Why not? I use Copilot to have an agent delegate to a different agent with a different model for certain tasks. Gemini is my technical writer.
copperx 2d ago on HN
I do it often, in fact, I just did it with Neon, but if Gemini is taking the output of Claude to extend it or summarize it, it gets "infected" and the resulting prose is Claudified. It's a gnarly issue at the moment.
aktau 2d ago on HN
Subjective.

In my opinion, the Argon class models (Gemini 4 "pro", so not Hydrogen). have the best writing voice.

It's still LLM-ese mind. But it's easier (possible?) to prompt it into some type of submission. Claude is very resistant to prompts telling it (how) to write better.

weitendorf 3d ago on HN
Anthropic doesn't want people to use their models for the chat/persona style use cases OpenAI has spent more time and effort optimizing for.

If you look back at GPT2 and the early days of AI safety/risk this was actually something the kinds of AI researchers that joined Anthropic didn't want - RL on coding->using that accelerate development and to bootstrap capabilities across domains was always the plan, not making it easier for people to spam/fake homework or trap people in problematic human-AI attention/relationship tarpits.

I personally find that Opus 5.5 and Fable 5+ are generally fine to chat with but just very wordy and culturally "LessWrong/essay-guy". Which, if a model is going to have a personality, is close to the one that I want (though it does piss me off when it gets preachy about standards Anthropic clearly doesn't hold themselves against)

I also want to immediately recognize AI generated content on the web or from another person, so I'm very glad it has a distinct voice

quixoticelixer- 3d ago on HN
Yeah maybe that's the case, and maybe I am more sensitive than other people. But I don't really use AI for that sort of thing but even for technical stuff I still find it's writing style a real turn off. Gemini still has lots of AIisms but it's a lot more tolerable.
Degorath 3d ago on HN
Have you tried Muse? I only took it for a month on a cheap plan and I did not do much serious work with it, but my initial impressions on the writing style when I wanted it to explain me things was pretty nice.
quixoticelixer- 2d ago on HN
No it's only just become available in my country
Yizahi 2d ago on HN
Discussion about issues with Anthropic LLMs: "Oh, they did that on purpose to thwart some imaginary future problem or legislation, they are genius like that."

Discussion about issues with Google LLMs: "Have you considered that Google are simply bad and failed to do that properly?" (saw this exact line at least a dozen times now on HN)

LeoPanthera 3d ago on HN
If the training set is smaller, doesn't that make sense though? There will be much less variety in its style and so its going to make very recognizable output.
quixoticelixer- 3d ago on HN
Yeah that could be it. I guess the thing about what makes AI writing so repulsive sometimes is not that it's bad like a human would be bad, but that it does the same things over and over.
impulser_ 3d ago on HN
I don't think it pre-training, because surely these labs all have similar data and Google should be very good at ingesting, cleaning, and filtering data.

I honestly think it just their post-training is fantastic and I think it has something to do with them being very opinionated on how they want the model to act that it basically the best at roleplaying a human. Humans are very good at completely and figuring out things that there aren't "trained" on if given basic set of intelligence. So when it working on these long tasks that might not be directly in it training data it still reason through it very well.

They got made fun of for this in the pass because they do a lot of research on this including hiring a philosopher.

This is just my best guess, because you don't see any other labs talk about how they made the model act and Anthropic is basically obsessed with it and talks about it every model release.

I think they just have better "taste" and that taste is their secret sauce.

tarsinge 3d ago on HN
> Google should be very good at

Given all the other things Google should be good at and is failing at why would this be the exception?

Topfi 3d ago on HN
Purely outside looking in, but yes, I'd argue you are right and higher-quality is a major part of it, both for pre- and post-training. There appears to be a far greater willingness to adopt specifically tailored training sets from third-parties, consistently update and build on them.

From simple frontend work to strategically adding virtual rulers when working with LIDAR scans to extract more accurate data, there are very clear points when Anthropic introduces a new external data set into their mix and while I have seen the same from competitors such as Google, there simply appears to be a greater tendency to stick with these once they have them.

As an example, at release it was quite apparent that Gemini 3 Pro, 3 Flash and Opus 4.5 had a similar, possible even the same frontend-focused training base bought in from a third party. But while Anthropic has had both smaller and larger improvements across each subsequent Opus release, I have not noticed such a consistent improvement with Gemini models over the last year. Maybe Anthropic simply buys updates more often, maybe they create more artificial PRs on top, who knows.

Same with the LIDAR scans, there was a point a few months ago when models started to exhibit a very systematic way of handling visual references in some few leather-crafting tasks, enabling whole new ways of working more efficiently combining existing measurements with images. Step-by-step pixel measurements, compensation for angles and poor captures, etc. Have since seen this in other models, again very much feels like a bought in data set that they are pushing as far as it goes.

Then there is 3D capability with models like Muse Wonder Full. I have not used it myself, if it ain't on API, it doesn't exist, but have seen PRs and do find the model output very parsable, despite it being terrible at any modelling task because these capabilities do not emerge from just training on vast amounts of data, they require high-quality, well tailored, step-by-step reference data of the entire process.

Harness wise, while pure output performance is very consistent across harnesses in my evals (I nowadays use both OpenCode via API and the labs own harness three times each to normalise), Anthropic also appears to be far more willing to use the time between releases to prepare features for the next. Stuff like workflows, proper subagent implementation, advisor, etc. often have found there way into Claude Code either in-between models or were first seen with a model that couldn't fully leverage these. Also feel that explained the good overall launch experience Fable 5 delivered, they simply used the time between announcing Glasswing and launching full access to get the UX/DX polished.

Meanwhile, I am still testing GPT-6.2 Astra and Sol so will refrain from making any more judgement on either till I have all data, but apparently there has been an extensive last minute rewriting of the system prompt to improve performance and "behaviour". Given the gap between GPT-6 Astra and yesterday, along with statements by OpenAI staff alluding that they had this model ready for a while, such things should happen in a far more structured, systematic, well thought out manner, not last second after some negative outside feedback. Heck, from what I know Anthropic, unlike any other lab, does zero early access outside the limited hardening for very exclusive partners like Glasswing. Certainly no early access for regular users, reviewers or the like from what I've heard, so Anthropic does all that prep-work for model releases with less feedback than other labs.

hackingonempty 3d ago on HN

  Anthropic spent a lot more time editing their pre-training data with higher-quality material. For example, they used actual books for pre-training, because books are generally run through human editors for higher quality. (this is from the copyright lawsuits)
  [ ... ]
  Meanwhile its obvious that companies like Google would use trash data like email and chat logs or Internet forums for pre-training. Just complete trash data.
Google has been scanning every book they can get their hands on for 20 years or more. They have a bunch of patents on non-destructive book scanning gear. Surely they are using that data to train their models.
wahnfrieden 3d ago on HN
Surely they are destroying scanned books for training
mastax 3d ago on HN
They invented those nondestructive scanners before the court cases that effectively mandated destruction.
wahnfrieden 3d ago on HN
Why does that allow them to use the scans for LLM training? Those scans were used for a different purpose. I don’t understand the logic.
yard2010 3d ago on HN
The logic is: when you have a lot of money, you get a lot of influence and you get the freedom to choose what to do. Be it traffic humans, induce riots, steal property, etc. Oversimplified, but this logic holds most of the time in my experience.
crimsoneer 2d ago on HN
No, the logic is that if you have a copy of a book, then create a copy and destroy it, you still have a copy of a book (and so can't go and sell on the original).
Schlagbohrer 3d ago on HN
If I recall correctly its because legally they (Anthropic) can avoid a copyright lawsuit if they destroy a rare book, because digitizing it and then having the original no longer exist in any other form is somehow... it prevents any possibility of a lawsuit due to a quirk in the law.
hackingonempty 2d ago on HN
While Anthropic destroyed the books and the fact that the new copies substituted for the old ones was a factor in the fair use analysis, it may not be necessary. Google already prevailed in the case brought against them by the Authors guild for nondestructively scanning for building a search index and only showing snippets to searchers.

https://en.wikipedia.org/wiki/Authors_Guild,_Inc._v._Google,....

foohis 3d ago on HN
won't help if they also include trash internet text in the training set as well.
etoxin 3d ago on HN
it's kinda known that the gemini models write better english which goes against the "trash data" remarks.
Gareth321 3d ago on HN
Maybe not. If they’re trained on real world emails and Reddit comments, they’ll write more like “real” people, but they’ll be dumber.
rixed 3d ago on HN
Exactly my gut feeling as well.

It's not that Claudese is wrong, it's that it's so dense as to become enigmatic.

perfsakuya 2d ago on HN
Enigmatic language increases the cognitive load on human readers. In this regard, maybe Gemini is better.
foohis 3d ago on HN
Have you ever tried the Gemini models? They are absolutely atrocious. Like, hilariously bad. One time they kept asking me for "Harriet! Where is Harriet?!" and another time it responded in chinese.

They were not trained on good data.

And higher-quality source material results in higher-quality emergent behavior, everything from thinking to tool use.

Seriously people, read their copyright court case. It literally states why Anthropic focused on books and higher quality material.

Google has to learn to not use trash internet content if they want to catch up.

(note that I haven't tried Neon yet so maybe they learned their lesson and fixed it. All my experience is from 3 months on their Ultra plan and previous models from earlier this year)

nl 3d ago on HN
This is incorrect.

Everyone uses books as well as internet data.

Anthropics difference is they have a huge data budget for buying RL environments for real world tasks.

positr0n 3d ago on HN
I get recruiter emails for $150/hr to do code review training for anthropic (W-2, although short duration contract).
nl 3d ago on HN
Yes, and if you look on LinkedIn there are many companies doing similar tasks in areas from accounting to physics to HR. They sell to all companies, but Anthropic seems the most aggressive. This from last year:

> According to The Information, leaders at Anthropic have discussed spending more than $1 billion on RL environments over the next year.

https://techcrunch.com/2025/09/21/silicon-valley-bets-big-on...

jonah 3d ago on HN
For other fields like medical or legal it's 2-3x that.
foohis 3d ago on HN
They do have a huge RL budget, but their OG primary source were all books they scanned. Books were a lot cheaper and higher quality than other data. This is straight from their court case.
nl 2d ago on HN
Everyone else used the same books.

OpenAI: https://news.bloomberglaw.com/ip-law/openai-risks-billions-a...

Google: https://www.theguardian.com/books/2026/jul/14/publishers-sue...

There is zero evidence that book data significantly increased coding performance. It increases coverage of knowledge, but not much else.

UltraSane 3d ago on HN
I think by now everyone training trillion parameter models is using almost every book, magazine, research paper, etc they can get.
mlmonkey 3d ago on HN
Microsoft (MSAI) showed that quality >> quantity with their Phi model a couple of years ago. It was a small model, but trained on very high quality data, and did better than much larger models.
nl 2d ago on HN
Phi was massively overfitted on the data and didn't generalize well in practical use though.
amarcheschi 3d ago on HN
I can't say what precisely, but other big companies are also doing what they can to get human data of high quality, just look for all the companies that popped like mushrooms that do ai reinforcement tasks, micro1, ethos, mercor ecc
olalonde 3d ago on HN
Where do you get that information from? Or is it just speculation?
mrheosuper 3d ago on HN
Is that reason why Claude talks so like .... a nerd.
potamic 3d ago on HN
Is this your opinion or you have a source for your claim?
foohis 3d ago on HN
Not opinion at all. It's directly from their court case.

"Over time, Anthropic came to value most highly for its data mixes books like the ones Authors had written, and it valued them because of the creative expressions they contained. Claude’s customers wanted Claude to write as accurately and as compellingly as Authors. So, it was best to train the LLMs underlying Claude on works just like the ones Authors had written, with well-curated facts, well-organized analyses, and captivating fictional narratives — above all with “good writing” of the kind “an editor would approve of” (Opp. Exh. 3 at -03433). Anthropic could have trained its LLMs without using such books or any books at all. That would have required spending more on, say, staff writers to create competing exemplars of good writing, engineers to revise bad exemplars into better ones, energy bills to power more rounds of training and fine-tuning, and so on. Having canonical texts to draw upon helped"

From: https://copyrightalliance.org/wp-content/uploads/2025/06/Bar...

nikhilsimha 3d ago on HN
i saw a researcher at uw claim that claude is trained on genome sequences and that others arent.

genome sequences are probably very dense in “reasoning”

oasisbob 3d ago on HN
I have a weak theory that some of the strange contortions in Claude come from training on scientific literature.
__s 2d ago on HN
Was recently trying to read "categorization is baked into the brain" as someone who doesn't read papers it tickled my brain similar to the hours of slop I read per day
MattRix 2d ago on HN
I don't think this is true because Opus 5 was bad and Opus 5.5 was really good, and presumably those both used the same pre-train? So instead it's more of a post-training thing?
thanhhaimai 3d ago on HN
It's likely that the agent configs you have might overfit to Claude models.

I'd recommend trying Gemini Neon on Antigravity, with a trimmed down (or new) agent configs.

svara 3d ago on HN
I think Gemini 4 (Argon is the one I've been using) is very clearly better at writing English than any Anthropic model. It's not better otherwise but in that the difference is vast.

To the point that I now use a workflow of delegating from Opus or Fable to Gemini4 to write commit messages and PR descriptions, it mostly solves the Claudese problem.

behnamoh 3d ago on HN
> To the point that I now use a workflow of delegating from Opus or Fable to Gemini4 to write commit messages and PR descriptions, it mostly solves the Claudese problem.

how do you do this w/o breaking bank? for Argon to write good commit messages, it must read the diffs and all the work Opus has done, which could easily eat up its usage quota.

rolosa 3d ago on HN
Can you just have opus hand off the changelog to Gemini and have it write from that?
spolitry 3d ago on HN
You vastly underestimate size of context the model can process.
lilbigdoot 3d ago on HN
I use a $20 a month plan with Argon and I honestly find it pretty usable for asking a lot of questions if I don't keep it all in one session, or it can do a few small coding tasks a day for me until I hit my limit. I haven't hit any limits with the smaller models cuz I don't like using a lot of LLM code in my personal projects since me learning a new thing or area by coding is one of the main motivations. Generating messages from diffs/notes after the work is done is pretty cheap I imagine.
relistan 3d ago on HN
That would be expensive, but there’s a simpler way. First though: I don’t use Argon for that. The cheaper Gemini models are quite good writers, too.

Second: you don’t need to waste a lot of tokens having the other model re-read commits and code. My AGENTS.md/CLAUDE.md specifies to write PR summaries and titles and then hand them off to Gemini to clean them up. Very occasionally this requires two trips if Gemini changes something the other model disagrees with, but usually results in very clear and readable PRs, often better than I would have written if I spent 15 minutes on it. And it uses very few tokens.

hikarudo 3d ago on HN
I just use Opus with the simple-english skill. No more Claudese.
NikhilVerma 3d ago on HN
I've been maintaining SlopSift - https://slopsift.dev/ for a while it's powered by a dependency parser to deterministically lint English. Then I wrap it around Claude either as a skill or a hook. It's really simplified the AI speak for me
dgfl 3d ago on HN
And yet the prose in the website is still full of claudisms. I’ve been looking for a tool to improve prose, but I still haven’t found one that obviously works. I’m starting to think Claude is just unable to output a sentence that doesn’t sound like itself. Have you run it on your own homepage?
spider-mario 2d ago on HN
Yeah, I’m curious what it would say about:

> Authorable rules inspect the graph for structural tells. Every finding names what matched and the exact text that triggered it.

(I don’t know if it’s just me but those are also very much garden-path sentences for me.)

rspeele 2d ago on HN
I think what makes de-slopping text so hard is that it starts with what the AI chooses to talk about. That cannot be corrected in a cosmetic pass. You can play with the word choice and sentence structure, but the statement in your quote will always read as Claudish, because Claude loves to proudly state that some rule was followed on every instance of a thing.

I personally hope this never gets solved so the valuable signal of “the person sharing this thing cared enough to describe it themselves” remains detectable.

Degorath 3d ago on HN
Could you link it, if it's open? I haven't had good success with ASD-STE100 myself :(
hikarudo 2d ago on HN
Degorath 2d ago on HN
Thanks, tried it, but it did not really solve the issues I have with claude :(
relistan 3d ago on HN
This is better, but you might try the Gemini models if you have access. They are much better writers, even with the simple English skill.
aloha2436 2d ago on HN
Simple English skills get rid of the most egregious word choices but their awkward sentence structure still comes through for me. I can definitely still tell, if only because I know what my coworker's actual writing sounds like.
ignoramous 2d ago on HN
> their awkward sentence structure still comes through for me

My fear is that folks who read Claudish all day everyday will start picking up its style.

losvedir 2d ago on HN
Nah, I haven't found complete success with tweaks like that. I would strongly suspect they're still there and you might just not be as sensitive to them.

I would bet you still get tells like overusing "never" or ubiquitous inanimate subjects with animate verbs ("questions that decide the option").

disillusioned 3d ago on HN
Gemini has reliably been the best writer of the bunch, and Argon has blown me away with how "human" it can sound. It has so much less of the AI smell than the others, though Opus 5.5 definitely lobotomized the absolute sloppery that was Opus 5, at least.
antupis 3d ago on HN
Yeah usually I use Gemini for writing. Anthropic and OpenAI models even with heavy prompting sound like LLMs where Gemini is way more natural sounding.
relistan 3d ago on HN
Hydrogen is also quite a good writer, is much faster than Argon, is cheaper, and for most things is very good. Currently one of my favorite models all around.
relistan 3d ago on HN
Yes! All the Gemini models are better writers by a lot. I also delegate PR summaries and titles to Gemini from the other models. Even when you add extra effort to get Claude to humanize, it is still nowhere near as good as Gemini.

For planning documents about complex work, Hydrogen does a fantastic job of both the analysis and the writing.

cromka 2d ago on HN
Do you delegate by hand or thru a skill or some general instruction in your AGENTS.md?
relistan 2d ago on HN
I use an AGENTS.md instruction
paganel 2d ago on HN
> from Opus or Fable to Gemini4 to write commit messages and PR descriptions

Who's reading that slop? Humans?

cowlevel 2d ago on HN
LLMs can read commit messages.
paganel 2d ago on HN
And pass that information to whom? For what purpose? It's slop all the way down?
cowlevel 2d ago on HN
Themselves when writing more code
paganel 2d ago on HN
So it is indeed slop all the way down.
jjbinx007 2d ago on HN
I agree. Gemini writes excellent English and doesn't trigger my AI-dar but I can tell Claudish from a mile away. In fact, I just had a notification from Life360 and spotted a Claudism so I now have a strong suspicion they're using Claude to assist with code.
HSO 2d ago on HN
aidar )))

im gonna steal that

qalmakka 2d ago on HN
Agree. Claude sucks at writing, you can pinpoint it's Claude from a million kilometres away and it always sounds like it's writing in slogans or something. Both GPT and Gemini are way better at writing text
m_ke 3d ago on HN
Anthropic has a proper feedback loop with claude code. They see way more real production data and can detect where their models are weak.

So they can train on prod traces or at the very least use user prompts to seed good synthetic tasks to train on.

chilmers 3d ago on HN
OpenAI has plenty of feedback from Codex. Maybe not quite on the same volume as Claude Codex, but still massive. So I'm not sure that's the whole story.

I doubt there's a single reason for why Anthropic are ahead right now. Building a frontier model requires executing well in a wide variety of things. It's probably a result of being slightly ahead in various aspects of of data collection, model architecture, pre and post training, harness development, etc.

It's a shame that so much of what's happening is being closed doors. Maybe one day, decades from now, somebody (or some AI) will write a definitive history of this period, explaining what was going on at the various labs, how their research compared, and what the major players really thought.

cmrx64 2d ago on HN
the businesspeople more prone to using codex are liable to avoid contribute their data to training. whereas claude has endless amounts of enthusiasts.
risyachka 3d ago on HN
Results usually reflect company culture and with big companies no matter what you do - if the culture is not there - you are doomed.

Google is the best example of this with their product graveyard. No matter what they release- they just can’t make a good/nice/usable product.

Just check google ai studio as one of latest examples.

spolitry 3d ago on HN
AI studio isn't great, but what is a better lightweight studio product than AI studio?
risyachka 2d ago on HN
Isn't great?

That is a massive, massive compliment. You literally need a phd there to set up billing so they can take your money.

Vibe coders make more usable products. There is absolutely no excuse or reason to put products of that quality out in the open. But google keeps doing it over and over.

Which just proves - you can't change company culture. So new startups always have a chance as incumbents just can't change due to "laws of physics".

>> what is a better lightweight studio product than AI studio? A vibe-coded clone is better.

irthomasthomas 3d ago on HN
This model matches Fable 5.2 for about 40% the cost. Not sure how you can take from this that anthropic has some secret sauce?
cloverich 3d ago on HN
According to benchmarks; time will tell. I see a lot of comments here and on X about how Google's caught up. But when Opus 5.5 came out, I didn't see comments. I saw examples of what people built, and it was mind blowing. Stopped the world for a few days for me and others, while we re-configured our brains around what was possible. I'm entirely willing to believe Google's catching/caught up but - lets have some real people build and show off like they did with Opus and Astra.
DashAnimal 3d ago on HN
e12e 3d ago on HN
How do you evaluate opus 5.5 as "better"? My experience has been that Argon was ok - maybe equivalent to opus - but carbon in orchestration mode was finally able to rewrite some early experimental (ai slop) front end code - and break a bit out of UX/UI rut that a couple of projects were stuck in.

It might be that opus is better in better structured projects, and carbon is better at (re)structuring poorly structured projects?

Carbon is a better project manager/product owner - opus a better programmer?

porphyra 3d ago on HN
They just have an incredible data flywheel I think. Since they were first to market with claude code, they have more training data than others. They also pay a lot ($150 per hour) for SWE experts to train the AI. I get some random recruiting emails occasionally from companies affiliated with them.
jbonatakis 3d ago on HN
Yeah I’ve gotten the same thing. If anyone from Anthropic is reading this: it’s not enough to make me quit my job for a short term contract. Try doubling it and we’ll talk
mattlondon 3d ago on HN
I think all the big labs do this - the Scale AI company that Meta paid squillions for was basically a company that recruits experts for human training for the big labs.
danduma 3d ago on HN
You are not alone. I have the exact same experience: opus just gets it and the others feel like cheap toys
atemerev 3d ago on HN
The secret sauce is sandwich statements. Just look at any Claude response. Positive-negative-positive. Bamm, and we are hooked.
neximo64 3d ago on HN
Exclusivity on RL environments they license.
sudohalt 3d ago on HN
There isn't because Opus 5.5 isn't better. There is no objective reason to believe Opus 5.5 is better.
thatxliner 3d ago on HN
> Is there any theory on what is Anthropic's secret sauce?

I'm planning to write a blog post on it, but my theory is that it's aligned to not be "politically correct" but instead to have a solid stance. In the Claude Constitution, they explicitly say how they train it to have judgement and not "follow rules"

antupis 3d ago on HN
It depends for programming Opus/Fable is the best but eg for code reviews or documentation Carbon is better. I think here Google own code review/documentation practices and thus probably best dataset in whole world really pays dividends.
cavisne 3d ago on HN
My theory is firstly its focus - they got to skip to GPT 4 (ie clean room implementation by the same people), they never cared about consumer or video gen or image gen.

Secondly, they somehow made all this work on public clouds with really crappy (compared to baremetal nvidia) networking. I think thats probably a lot of post training (which doesn't need such a large/tightly connected cluster), and some architecture that is less reliant on a tightly connected cluster for pretraining. This would also explain why they have been able to use Trainium, and somehow beat Google while using TPU's.

socalgal2 3d ago on HN
I'm personally having amazing luck with Gemini Argon at work (what our company chose). I use Claude at home and have had great luck there too. But, they are much more complicated issues at work than at home so I don't really know how to compare them. I tried a few simple graph one-shot demos and Claude (Opus 5.5 - medium) did better than Gemini Argon medium for those small things. But for big things on our company's C++ code base, Gemini Argon has been amazing. I don't have Claude there to compare.
DetroitThrow 3d ago on HN
Argon seems really good at C++ in particular in my experience. I prefer Claude for Rust, though.

They both seem to break down at similar size codebases or complexity levels, with maybe Opus having a slight edge over Argon.

BetterThanSober 3d ago on HN
> Anthropic's secret sauce

For casual use it's the harness for me, using Gemini's web interface feels like I'm fighting with all the options while on Claude's I just shoot and let the AI activates whatever tooling they need

scotty79 3d ago on HN
> Opus feels like human level intelligent like I'm talking to a coworker basically while GPT and Gemini feel like I'm talking to an AI and have to change how I talk.

It's possible that after 4o OpenAI has RL to make its models less human.

tekawade 3d ago on HN
I mostly use AI with coding. IMP Claude Code is the secret sauce. It had pioneered and now continues to evolve many new concepts. Which are later adopted by others.

Maybe someone used all with API and goose or open code can chime in here.

mattlondon 3d ago on HN
For what it is worth, I find Opus really, really irritating to use. Just the way it talks is really annoying.

If they've done anything with their special sauce, it appears to be to train it on the behaviours of jerk engineers who are smarmy, overconfident, and arrogant.

I prefer to use the Gemini models - if I need to spend 8 hours a day talking to something, I dont want that something to be sanctimonious Opus models.

nbardy 3d ago on HN
I think a lot of it boils down to "constitutional AI".

They did a paper on this a long time ago[1], I'm sure they have altered their techniques some, but I this was the earlier signal that they really treated the AI as a separate entity and trained it as such. And focus not just on it's capability, but it's *behavior*, when to push back on the user, etc... Imo. it's more about the philosophy of the constitution in this paper than the specific RLAIHF, I'm sure this idea of "when to follow instructions and when to trust yourself is in their system prompts for all their RL environments, and even their rewards.

This idea of an objective function vs what the model should do during human interaction is a huge question and a very subtle one to get right, but it's why they were winning in real world application even before they were winning in raw benchmarks.

Especially in the era of RL environments dominating more and more compute it's vey easy to overfit a model to be very good at accomplishing a narrow task exactly as specified. Much more complicated to have it respond appropriately on turn 5 of a nuanced conversation with a user about their private codebase.

[1] https://arxiv.org/abs/2212.08073

charm137 3d ago on HN
Their models simply overthink - they are simply some of the most token inefficient models from the frontier labs: https://artificialanalysis.ai/?intelligence-efficiency=outpu...
mrklol 2d ago on HN
If that’s the current way to get the best results, I wouldn’t see that as a problem
charm137 2d ago on HN
It isn't the best way, as is visible from the Pareto frontier curves here for GPT.

Secondly, this strategy leads to Claude models also being the slowest models - so each response / tweak iteration wastes more time than other models.

Thirdly, because Claude's intelligence drops off so rapidly with thinking levels, you end up paying way too much for marginally higher intelligence (which may not even be perceptible) for its Max levels whereas the equivalent levels of its competitors cost less due to fewer output tokens being used (and generally saner pricing).

gallerdude 2d ago on HN
It doesn't matter if their car gets 10 miles per gallon if it's the only car that can drive faster than 30MPH.
charm137 2d ago on HN
That analogy inadequately addresses the reality and I respectfully disagree.
vintermann 2d ago on HN
There are a lot of real world specialized tasks you can't get better at by logical reasoning or role playing as a human. How you conjugate in some small southeast Asian language. What handwriting looked like in 18th century Norway. These are the tasks which first degrade when you shrink a model. Google's models, in my experience, do well on them, other models not so much.

It makes sense too, because of course Google should have a ton of exotic training data. It's probably their comparative advantage, over getting super-clever with reinforcement learning.

ozgung 2d ago on HN
> Is there any theory on what is Anthropic's secret sauce?

Yes, it's called the "Soul Document". A long trade secret document depicting everything about Claude (behavior, values, character, boundaries etc.) in great detail. They use that in training to shape Claude's identity.

I think this makes all the difference because I feel like talking a very good coworker. It understands me, helps me and slightly pushes me to be a better version of myself. This is in contrast with a robot blindly doing what I want.

bbstats 2d ago on HN
I think tool calling is a good % of this
WhereIsTheTruth 2d ago on HN
Continuity
dbolgheroni 2d ago on HN
Did you use the same harness/agent for both?
troglodytetrain 2d ago on HN
Dedicated RL for their harness + dedicated RL for 'orchestration' as their special focus.

Only evidence I can point to is that over a year ago I built a custom ai harness for a corp with no usage caps on anthropic models. Claude taught me all the buzz words as I built long before they were publicly discussed anywhere. From that period try searching for "Orchestrator" hits on google/github etc. My code has it for the agent with the subagent calling toolset, Claude's recommendation. It performed long horizon tasks really well too. Almost exactly like Fable, same persistence, with worse performance of course/accuracy/hallucination. Just had a watchdog script and a validation agent that ran when the main orch claimed it was done and yell at it about what it didn't complete, all on 250k context sonnet.

Now, bundle that basic format with a nice conversational LLM that Anthropic keeps pretending is its own model doing everything by itself with its much bigger context and you get what we have today.

Essentially providers specialize their AI stack + LLM personality + backend fuckery they pretend isn't 90% of model capability into their api spigot. Openai has been focused on 'better chatbot' for the most part and Google appears to have been sitting back and waiting to see what worked first.

Then, quickly with the big perf upgrade with the 'orchestration layer' path, Google had a target to quickly RL on. Thats why you should only be using Agy on Carbon with /Swarm on IMO. Its great. Carbon is fine tuned for orch work. Same reason Fable often performs worse at benchmark tasks than Opus. Its meant to be an Orch not a worker.

dennysora-main 3d ago on HN
This is good, it makes me have to understand the full names of every element in the periodic table.\ (PS. Taiwanese)

But I still can't understand the differences between these models.

Recently I watched Veritasium talking about helium, and I was thinking, will the effort setting of these models eventually have something like "Temperature K" to configure?

ncd 3d ago on HN
Okay, but now will you let my tiny team sign up somewhere to be able to use this the way I use `claude` or `codex`? I'm told to use `antigravity`, but when I go to https://one.google.com/ai (which is where the magic answer from Google search told me to go), I'm told that I have to use my "personal" account, not my work account. So I clicked "Manage Google Workspace Account" on that page, and got dumped in a super dense UI with nothing related to enabling this model somewhere.

Good lord, this is an insane flow.

smallerize 3d ago on HN
So I haven't tried this, but it looks like you have to set up the Gemini API in a Google Cloud project, then point your Antigravity at the project. https://antigravity.google/docs/pay-as-you-go/
ncd 3d ago on HN
Thanks for finding that. Looks like it would use API pricing? I was hoping to be able to get a subscription like offered by the other competing models, but might not be able to do that.
dekhn 3d ago on HN
Waiting for Gemini 4 Oganesson
nkmnz 3d ago on HN
Waiting for Gemini Plutonium to launch.
horseloverthin 3d ago on HN
I think that its likely the reason for a) the strange naming scheme b) the promptness between releases is that google has outpaced internal expectations for model development ( probably due to RSI ) and now can now utilize its vast compute pool much more effectively for autoresearch and exploratory self-improvement tasks. I wonder if this suggests a steeper trajectory for them than OAI/Anthropic in general.
xd1936 3d ago on HN
I have the "Google AI Pro 5TB" plan for Students. I see "4 Hydrogen" and "4 Neon" in the model picker on gemini.google.com. When I look at one.google.com's plan details, I see "Your Plan: Google AI Pro (5 TB)". When I click "View Benefits" on that page, one option in the list says:

Google Antigravity

Entry rate limits to advanced models, including Gemini 4 Argon and Opus 5.5, in Google Antigravity, our agentic development platform

Yet, here I am, running the `agy` cli, and the model picker says "Unavailable: Gemini 4 Carbon/Argon/Hydrogen/Neon is currently available to paid Pro and Ultra subscribers". It has been this way since the release of Carbon. I guess it's "3.8 Flash" or nothing for me.

mathis 3d ago on HN
Hydrogen, Argon and Neon are included in Pro. Carbon is not. At least for me. You might want to update your cli.
pzo 3d ago on HN
Not true. Hydrogen and Neon are available only in web chat. 3.8 Flash is the only one available in antigravity for such promo subscriptions.
rawgreaze 3d ago on HN
I had to grab my laptop to say, you are actually incorrect. I have been using Argon for the last week or two via Antigravity on the Google Pro $20/mo plan, discounted to $5/mo for 3 months, just fine. And as of today yes, I see Neon and Hydrogen there. The only one I can't see/use is Carbon. https://i.imgur.com/2c9EnbV.png. This isn't just a visual thing, I have actually used every model there. They all work.
xd1936 2d ago on HN
They said, _for such promo subscriptions_.

Said another way, I am using "Google AI Pro" free for students, which seems to be a deceptively different plan internally than "Google AI Pro" for consumers, even though they list the same features on their marketing websites.

anvuong 3d ago on HN
It's available as long as you pay something. The student 12-month Pro at $0/month only has 3.8 Flash in agy, but if you pump it up the $8.99/month (+youtube premium) then it becomes the regular Pro.
xnx 3d ago on HN
> Neon writes clearly, with well-structured responses and a natural, direct tone

This is an underrated quality of Gemini models.

Tadpole9181 3d ago on HN
Does anybody have a corpus of examples I could look at? I've wondered if it's worth setting up Gemini as a human-facing documentation author with other models still being the underlying workhorse.
hippycruncher22 3d ago on HN
Can someone explain wtf is going on at google?
mpvjs 3d ago on HN
gemini tungsten can
seizethecheese 3d ago on HN
It looks like it's being served extremely fast via Openrouter right now https://openrouter.ai/google/gemini-4-neon. 127 tokens per second. This is way faster than Carbon at 46 tokens per second (https://openrouter.ai/google/gemini-4-carbon). I'm a bit confused by this, since it seems like Neon is supposed to be their flagship model. Perhaps they've been throttling Carbon?

Carbon has a huge problem that it's just way too slow, and outputs way too much per task (on real world tasks, it's the worst at this https://micjm.com/blog/carbon-is-that-slow/).

Without the speed issue, Carbon would be a top model already. It's top on our battle system built on opencode, if you exclude early votes: https://agentduel.com/leaderboard?excludeEarlyVotes=true, though to be fair, small sample size so far.

Conclusion is that it's looking good for Neon. Even if it suffers from similar problems as Carbon, if it's even smarter and way faster, it should be a top model, maybe the best.

Edit: I was wrong about Neon being the larger model. I made this assumption because Neon has higher scores and they're both "v4".

Showing 400 of 765. Read the rest on Hacker News

Comments are loaded live from Hacker News and are not stored by Mid or Real.