Moonshot AI has released the full weights for its Kimi K3.1 Pro model, a 2.8 trillion parameter model with a 1 million token context window, on Hugging Face. This release allows developers to self-host the model, but requires significant computational resources, including multiple data-center GPUs. The model is a mixture-of-experts architecture with vision capabilities and is trained with quantization awareness. While the license is largely permissive, it includes scale conditions for large hosts, and the training data and code are not publicly available. AI
IMPACT Enables self-hosting of a large context window model, potentially lowering costs for high-volume users and offering leverage for API negotiations.
RANK_REASON Frontier-lab model release with system card.
- Hugging Face
- MoonshotAI
- Claude (Opus 5.5)
- GPT -6 Astra
- Kimi K3.1 Pro
- Kimi K3.1-Pro License
- Kimi K3.1 Pro Open Weights
- Moonshot AI
- vLLM
AI-generated summary · Google Gemini · from 4 sources. How we write summaries →