Thinking Machines releases Inkling-Small open-weights MoE model
Inkling-Small is a 276B-total, 12B-active Mixture-of-Experts model, about a quarter the size of Inkling (975B/41B), trained on NVIDIA GB300 NVL72 systems. It scores 31.6% on Humanity's Last Exam, beating Inkling's 29.7%, and is available on Hugging Face, Tinker, Together AI, and Vercel AI Gateway.
Featured · Mira Murati
How this story unfolded
3 weeks · 8 reports · 7 community posts · 15 of 16 shown
- Jul 30
- Jul 31
- Aug 2
- Aug 21
Thinking Machines by email
Get an email when Thinking Machines has news
No news that day, no email.
More stories today
- AutoResearchClaw generates research papers from chat prompts
- 63% of Religious Books on Amazon Likely AI-Written, Study Finds
- Why real-time AI at scale is so hard
- OpenAI users report ongoing usage limit issues
- Claude user shares simple skill to track long sessions