Thinking Machines ships Inkling 2 Small: 594B MoE, Apache 2.0 weights next week
pmttyji · reddit · 2026-11-06
Thinking Machines introduced Inkling 2 Small, a 594B-total/17B-active MoE model with image input and variable effort. It scores 31 on the Artificial Analysis Intelligence Index, surpassing prior models via more efficient pre-training and a large unified RL run. Available now on Tinker with closed weights; full weights will be released under Apache 2.0 next week after safety testing and red-teaming. The previous Inkling Small was only 276B/12B active.
Related event: Thinking Machines Releases Inkling 2 Small(52 posts)→
More from Models
- Dev still on Claude Opus 5.5: 'too good', but wishes for fewer refusals — ivan_bezdomny · 2026-11-06
- Prediction: LLM decoding goes compute-bound next year; Sol 6 may decode sparsely — teortaxesTex · 2026-11-06
- Insider claims OpenAI's internal model shows math 'research taste', RSI months away — coherence · 2026-11-06
- OpenAI's cyber model finds 111 Firefox renderer RCEs, 81 of them novel — Dimillian · 2026-11-06
- Quantifying "AI flavor": 520万字 corpus and 500k semantic judgments become a de-AI skill — sujingshen · 2026-11-06
- DeepSeek Harness desktop reportedly hits 5.48M MAU and 261 daily requests per user — teortaxesTex · 2026-11-06