Kimi Models
Kimi support is inference-first for now.
Inference
| Model | Entry | Method | Status |
|---|---|---|---|
| Kimi K2 Thinking | kt run kimi-k2-thinking or manual SGLang-KT launch | RAWINT4 | Current / Needs smoke. |
| Kimi K2.5 | Manual SGLang-KT tutorial | RAWINT4 | Manual / Needs smoke; not registry-first yet. |
Fine-Tuning
Kimi SFT is not current KT SFT support. Old Kimi SFT pages remain legacy or experimental notes until the current LLaMA-Factory path supports Kimi and passes smoke.
Source Material
Current inference material comes from:
doc/en/kt-kernel/Kimi-K2-Thinking-Native.mddoc/en/Kimi-K2.5.md
Old SFT material is not a current quick start.