webAI Releases TwIL-8B, an 8.8B Reasoning Model With 98.2 on MATH-500
webAI released TwIL-8B, an 8.8B reasoning model post-trained from IBM Granite 4.2 8B. It posts 98.2 on MATH-500 and 81.6 on LiveCodeBench, but trails similar-size models on knowledge benchmarks, according to MarkTechPost.
How TwIL-8B ranks now
| Benchmark | Score | Rank | Source |
|---|---|---|---|
| LiveCodeBench | 81.6% | 18 of 48 | marktechpost.com, Nov 2026 ↗ |
Rank is against every model in our database with a sourced score on that benchmark, each counted once with its best score.
webAI has released TwIL-8B, an 8.8B-parameter reasoning model post-trained from IBM Granite 4.2 8B. The model scores 98.2 on MATH-500 and 81.6 on LiveCodeBench, according to a MarkTechPost report published November 4, 2026.
Key facts
- Model: TwIL-8B
- Developer: webAI
- Parameters: 8.8B
- Base model: IBM Granite 4.2 8B (post-trained)
- Type: Reasoning model
- MATH-500: 98.2
- LiveCodeBench: 81.6
MarkTechPost reports that TwIL-8B trails similar-size models on knowledge benchmarks.
What this means
TwIL-8B pairs very high math and coding scores with weaker results on knowledge benchmarks. For builders evaluating it, the reported profile points to mathematical and code reasoning as its strengths, and to factual recall as the area where comparable models of similar size perform better. It is also a post-trained derivative of IBM Granite 4.2 8B, which shows how far targeted post-training can shift the strengths of an existing 8B-class base.
Related Articles
Ricursive releases Ripple-65B Preview, 65B dense open-weight model, scoring 80.5 on Terminal-Bench 2.1
Ricursive Intelligence released Ripple-65B Preview, a 65B dense open-weight LLM under Apache 2.0. It scores 80.5 on Terminal-Bench 2.1, ahead of US open models up to 15x its size, according to MarkTechPost.
Cognition launches SWE-2 Flash, claiming three times the speed and one-fifth the price of SWE-2
Cognition announced SWE-2 Flash on November 4, 2026, and reports that the coding model runs at three times the speed and one-fifth the price of its larger SWE-2 sibling. It is available in Devin Desktop and CLI, with free use through November 30, 2026.
ricursiveAI's 65B Ripple Preview launches on GMICloud: 16K context, $0.20/$0.60 per 1M tokens
Ripple-65B-Preview, a 65B-parameter model from ricursiveAI with native tool calling, is served in BF16 by GMICloud and listed on OpenRouter as of Nov 4, 2026. It has a 16K context window and costs $0.20 per 1M input tokens and $0.60 per 1M output tokens.
OpenAI's Images 2.6 lets ChatGPT generate animated GIFs with speed and frame editing
ChatGPT can now create animated GIFs from a text prompt or an uploaded photo, with controls for playback speed and individual frames. The feature ships with OpenAI's new Images 2.6 model, which OpenAI says also generates images faster. OpenAI says users now generate over 4 billion images per week across ChatGPT and its API.
Comments
Loading...