webAI-Official/TwIL-8B
text-generation · ♥ 226 · ↓ 22.7K
- Tasktext-generation
- Downloads8209
- Likes150
- Licenseother
text-generation · ♥ 226 · ↓ 22.7K
An 8.8B reasoning model for math, code and instruction following, post-trained from ibm-granite/granite-4.2-8b with full-parameter supervised fine-tuning and GRPO/DAPO reinforcement learning. It ships in two variants that share one architecture and tokenizer: balanced (the default) and verbose. See Variants.
Against open models of similar size and up to 3.5x larger, TwIL-8B posts the highest MATH-500 (98.2) and LiveCodeBench (81.6 †) scores in the comparison and leads Qwen3.5-9B on IFBench (+9.9 †). It is level with Qwen3.5-9B on contest math while using fewer tokens per problem. It trails on knowledge-heavy benchmarks: GPQA Diamond, MMLU-Pro † and HLE †.
Excerpt from the model card on Hugging Face ↗.
webAI-Official/TwIL-8B was first seen on Hugging Face · Trending models on Nov 6, 2026. Since then it has appeared in 53 checks on 3 different days. Its best position was #15 of 30, reached on Nov 8, 17:42 UTC. It is #15 right now, climbing from #17 to #15 over the last 6 checks.
Position on Hugging Face · Trending models at every check (lower is better).
Best rank per day (UTC).
🔥 as reported by Hugging Face.
Archived daily, weekly and monthly lists that include this entry.
Yes. As of Nov 8, 17:42 UTC it is #15, climbing from #17 to #15 over the last 6 checks.
#15 on Hugging Face · Trending models, on Nov 8, 17:42 UTC. Trendoday checks this list every 60 minutes.
It was first seen on Nov 6, 2026 and has been on the list on 3 days (53 checks) since.
It is tagged text-generation with transformers on Hugging Face and has 8209 downloads under the other license.