← The Polylog AI Intelligence Brief
Morning Edition · Tuesday, November 10, 2026Published at 1:14 AM EST · New York
Moonshot publishes Kimi K3.1 Pro weights as tiny Qwen-based models win narrow tasks
Kimi K3.1 Pro is reported as a 2.8-trillion-parameter mixture-of-experts model with 104 billion active parameters and a 1M-token context, and an 800-million-parameter fine-tune of Qwen3.5 is claimed to beat far larger models at generating Bash commands.
Moonshot AI has posted the weights of Kimi K3.1 Pro on Hugging Face. Reports put the model at 2.8 trillion total parameters, 104 billion active parameters, 896 experts and a 1,048,576-token context. The repository is about 1.56 terabytes. O…
Continue the AI Intelligence Brief
Track frontier labs, chips, export controls, model releases, regulation, and AI infrastructure.
- 5 AI intelligence signals a day
- Frontier labs, compute, and chips
- Model releases and AI infrastructure
- Source-grounded analysis with confidence labels
The Global Intelligence Brief stays free.
Part of a tracked trend
Chinese Open-Weight Models Emerge as the Non-US AI Stack
As Washington restricts foreign access to US frontier models, governments and enterprises cut off from American AI increasingly standardize on downloadable Chinese open-weight models, splitting the world into competing AI supply blocs rather than a single frontier.
Townsquare
Start a discussion in Townsquare.
More from this edition
- Anthropic ships Claude Fable 5.2 at a 20 percent lower price and keeps Mythos 5.2 behind vetted access
- Anthropic heads toward a Nasdaq listing as OpenAI routes ChatGPT Enterprise through AWS procurement
- OpenAI's Control Room can halt any registered research job within 35 seconds and fails closed
- Terence Tao calls for a "Math 2.0" after OpenAI's release of more than 700 machine-generated manuscripts
- Claude helped resolve the first 3D structure of a prokaryotic vault particle, Anthropic says
- Meta's SAM 3 and DINOv3 run inside Energy Department beamline analysis and a robotic wheelchair project
- NVIDIA uses CoRL to move simulation-trained robotics from locomotion to fine assembly
Comments
0No comments yet.