VQV Signal
According to a post on X by @rohanpaul_ai, webAI has released open weights for TwIL-8B, an 8.8B-parameter model built on IBM's Granite 4.2 8B. The quantized file is 4.98 GiB and is described as fitting on CPUs or modest GPUs. The post cites webAI's reported 81.6% score on LiveCodeBench and its claim of about 93% of the best frontier score on coding and 92% on instruction following.
According to a post on X by @rohanpaul_ai, webAI has released open weights for TwIL-8B, an 8.8B-parameter model built on IBM's Granite 4.2 8B. The quantized file is 4.98 GiB and is described as fitting on CPUs or modest GPUs. The post cites webAI's reported 81.6% score on LiveCodeBench and its claim of about 93% of the best frontier score on coding and 92% on instruction following.
AI-assisted summary based on the listed source.
webAI released open weights for TwIL-8B, an 8.8B model built on IBM's Granite 4.2 8B that fits in a 4.98 GiB quantized file for CPUs or modest GPUs. It reports 81.6% on LiveCodeBench and says it reaches about 93% of the best frontier score on coding and 92% on instruction following. Why it matters: Capable coding...
The release points to smaller models with strong coding and math ability that can run on hardware many people already own. Developers and small teams that rely on local or low-cost inference could test whether these self-reported results hold up in their own workloads. The next thing to watch is independent benchmarking of TwIL-8B against the frontier models it is compared with.
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.
