{"name":"EdgeRock","repo":"rockmatrix-ai/EdgeRock","url":"https://midorreal.com/project/rockmatrix-ai-edgerock","repository":"https://github.com/rockmatrix-ai/EdgeRock","description":"EdgeRock: Strong Reasoning at Constant Memory. Multimodal models with O(N) generation and O(1) persistent history. Inference and evaluation code.","facts":{"license":{"value":"Apache-2.0","source":"https://github.com/rockmatrix-ai/EdgeRock","checked":"22 Nov 2026"},"language":{"value":"Python","source":"https://github.com/rockmatrix-ai/EdgeRock","checked":"22 Nov 2026"},"homepage":{"value":"https://huggingface.co/rockmatrix-ai/EdgeRock","source":"https://huggingface.co/rockmatrix-ai/EdgeRock","checked":"22 Nov 2026"},"releases":{"value":"0","source":"https://github.com/rockmatrix-ai/EdgeRock/releases","checked":"21 Nov 2026"},"contributors":{"value":"2","source":"https://github.com/rockmatrix-ai/EdgeRock/graphs/contributors","checked":"21 Nov 2026"},"archived":{"value":"no","source":"https://github.com/rockmatrix-ai/EdgeRock","checked":"22 Nov 2026"}},"stars":1237.0,"writeup":{"what_it_is":"EdgeRock is a family of multimodal models with fixed-memory persistent history. The first release, EdgeRock-4B, takes text and image input and generates text for reasoning, coding, and agent workflows. The repository provides inference and evaluation code.","audience":"Developers and researchers running reasoning and multimodal models with constant memory use.","claims":[{"kind":"specific","claim":"Uses a fixed 72 MiB of attention KV per sequence, 99.1% less than Qwen3.5-4B","excerpt":"fixed **72 MiB** attention KV per sequence—**99.1% less** than Qwen3.5-4B","status":"not_checked"},{"kind":"specific","claim":"Reaches 6.42× Qwen3.5-4B's output throughput at each model's best measured batch","excerpt":"**6.42×** Qwen3.5-4B's output throughput at each model's best measured batch","status":"not_checked"},{"kind":"specific","claim":"Leads 24 of 29 text benchmarks among bounded-state models","excerpt":"It leads **24 of 29 text benchmarks**","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-10-29"},"why_now":null,"owner_supplied":null,"score":{"verdict":"too_early","hype":72,"reality":15,"gap":57,"momentum":0.8,"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":18,"gap_high":58,"rule":"early-2.0"},"changes":[{"date":"2026-10-07T17:21:11+00:00","text":"Repository created","source":"https://github.com/rockmatrix-ai/EdgeRock"}],"cite":"EdgeRock is a GitHub project at https://github.com/rockmatrix-ai/EdgeRock written mainly in Python, described there as \"EdgeRock: Strong Reasoning at Constant Memory. Multimodal models with O(N) generation and O(1) persistent history. Inference and evaluation code\".","attribution":"Data from Mid or Real, https://midorreal.com/project/rockmatrix-ai-edgerock, checked 22 November 2026. Attribution with a link is required.","still_real":null}