Kimi K3.1 Pro weights released on Hugging Face under custom licence
Moonshot met its Nov 4 deadline: 2.8T parameters, 1.56 TB of MXFP4 weights.
Moonshot AI's full weights for Kimi K3.1 Pro went public on Hugging Face on November 4, 2026, the deadline it set when the model launched on the Kimi API on October 23. The repository is moonshotai/Kimi-K3.1-Pro, and it is not gated.
What changed
- The repository holds 96 safetensors shards, about 1.56 TB in total.
- Kimi K3.1 Pro is a mixture-of-experts model with 2.8T total parameters and 104B active per token (896 experts, 16 selected per token), a 1,048,576-token context window, and text and image input.
- The weights are MXFP4 with MXFP8 activations. Moonshot's model card says the model was trained that way (quantization-aware training).
- Compared with Kimi K3, Moonshot replaced the 24 full-attention Gated MLA layers with Gated DeepSeek Sparse Attention layers for long-context efficiency.
- Moonshot recommends vLLM, SGLang or TokenSpeed to serve it. vLLM announced day-0 support on November 4, 2026.
Licence terms
The weights ship under the Kimi K3.1-Pro License, not Apache 2.0 or MIT. It allows use, modification, redistribution and commercial use, with two conditions:
- A company running a "Model as a Service" business (API access to inference or fine-tuning) with more than $20 million in revenue over any 12 months needs a separate agreement with Moonshot before commercial use.
- A product with more than 100 million monthly active users, or more than $20 million in monthly revenue, must show "Kimi K3.1-Pro" prominently in its interface.
Neither condition applies to internal use, or to use through Moonshot's official products or certified inference partners.
Where to run it
- Moonshot's API, model ID
kimi-k3.1-pro. Prices are in our launch post. - Verda launched Verda Inference in preview on November 4, 2026, with Kimi K3.1 Pro as its first model (model ID
kimi-k3-1-pro). - Self-hosting: plan for about 1.56 TB of weights before any KV cache.
On RouterPlus
Kimi K3.1 Pro is not in the RouterPlus catalog as of November 4, 2026. Kimi K3, the model it was post-trained from, is available on 9 providers. RouterPlus catalog, November 4, 2026, per million tokens:
| Provider | Input | Output |
|---|---|---|
| Wafer | $0.60 | $15.00 |
| Sail Research | $1.50 | $10.00 |
| Chutes | $1.85 | $9.50 |
| Morph | $2.50 | $14.00 |
| DeepInfra | $2.85 | $14.25 |
| Nebius | $3.00 | $15.00 |
| Novita AI | $3.00 | $15.00 |
| Inference.net | $3.00 | $15.00 |
| Doubleword | $3.00 | $15.00 |
Lowest input is $0.60 and lowest output is $9.50.
curl https://api.routerplus.com/v1/chat/completions \
-H "Authorization: Bearer $ROUTERPLUS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "moonshotai/kimi-k3", "messages": [{"role": "user", "content": "Hello"}]}'Every provider and price: Kimi K3 API pricing. All Moonshot AI models.