{"name":"GVS5H","repo":"slee-persis/GVS5H","url":"https://midorreal.com/project/slee-persis-gvs5h","repository":"https://github.com/slee-persis/GVS5H","description":"GVS5H: Five Qwen3.8-27B Models Match Claude Fable 5 on LiveCodeBench Hard  |  Fable 5 Level Coding for a Fifth the Price - or on a Single GPU","facts":{"license":{"value":"MIT","source":"https://github.com/slee-persis/GVS5H","checked":"21 Nov 2026"},"language":{"value":"Python","source":"https://github.com/slee-persis/GVS5H","checked":"21 Nov 2026"},"releases":{"value":"0","source":"https://github.com/slee-persis/GVS5H/releases","checked":"19 Nov 2026"},"contributors":{"value":"2","source":"https://github.com/slee-persis/GVS5H/graphs/contributors","checked":"19 Nov 2026"},"archived":{"value":"no","source":"https://github.com/slee-persis/GVS5H","checked":"21 Nov 2026"}},"stars":1211.0,"writeup":{"what_it_is":"GVS5H is a research repository for a training-free method called ledger-based zero-shot self-orchestration. Fresh instances of one model decompose coding problems and coordinate through a shared filesystem holding a plan, notes and the current solution. The code runs a manager engine or a one-call baseline against LiveCodeBench problems.","audience":"Researchers and developers evaluating LLM coding benchmarks and multi-agent orchestration","claims":[{"kind":"specific","claim":"Orchestrated Qwen3.8 Flash Next reaches 93.0% pass@1 against Fable 5's 90.4% at 9% of the cost","excerpt":"Orchestrated Qwen3.8 Flash Next reaches 93.0% pass@1 against Fable 5's 90.4% at 9% of the cost","status":"not_checked"},{"kind":"specific","claim":"Locally served Qwen3.8-27B rises from 66.8% to 92.4% with the method","excerpt":"locally served, open-weight Qwen3.8-27B rises from 66.8% to 92.4%","status":"not_checked"},{"kind":"specific","claim":"Evaluated across eleven models on the 100 latest hard LiveCodeBench problems","excerpt":"Across eleven open and closed-weight models on the 100 latest hard LiveCodeBench problems","status":"not_checked"},{"kind":"specific","claim":"Gains reach up to 25.6 percentage points on pinned backends","excerpt":"the method yields gains of up to 25.6 percentage points on pinned backends","status":"not_checked"},{"kind":"specific","claim":"Gains are not universal, and some models are unchanged or worse","excerpt":"Gains are not universal: some models are unchanged or worse.","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-09"},"why_now":null,"owner_supplied":null,"score":{"verdict":"too_early","hype":87,"reality":13,"gap":74,"momentum":null,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T08:00:00+00:00","verdict_source":"rule"},"early_read":{"estimate":true,"lean":"too_close","gap_low":null,"gap_high":null,"rule":"early-2.0"},"changes":[{"date":"2026-08-26T20:56:52+00:00","text":"Repository created","source":"https://github.com/slee-persis/GVS5H"}],"cite":"GVS5H is a GitHub project at https://github.com/slee-persis/GVS5H written mainly in Python, described there as \"GVS5H: Five Qwen3.8-27B Models Match Claude Fable 5 on LiveCodeBench Hard  |  Fable 5 Level Coding for a Fifth the Price - or on a Single GPU\".","attribution":"Data from Mid or Real, https://midorreal.com/project/slee-persis-gvs5h, checked 22 November 2026. Attribution with a link is required.","still_real":null}