Kimi K3
2779.9B total, 105.8B active · 1024k context · Kimi K3 License · released 2026-06-13
chatcodingagenticvisionreasoning
Cheapest way to run it in the US today
Nothing in stock in the US today runs Kimi K3 fully in memory at 128k context at 15 tok/s or more. A shorter context needs less memory; the table below shows what comes closest.
What to run it on in the US
Fully in memory at 128k context, priced at retailers today.
Nothing we price holds this model fully in memory.
All hardware that runs it
Everything we price, best value first. Two cards split the model by layers.
Nothing we price can run it at this context.
Memory needed
Weights, KV cache and runtime buffers at each context length (KV cache at fp16).
| Quant | File |
|---|
Run it
Copy a command for the runtime you use.
No published builds to run yet.
Notes
Weights use native MXFP4 (QAT) with MXFP8 activations; thinking is always on and the full assistant message including reasoning_content must be passed back in multi-turn use. Recommended engines: vLLM, SGLang, TokenSpeed.
Questions
- What licence is Kimi K3 released under?
- Kimi K3 License.