Kimi-K2-Instruct
Support summary
Kimi-K2-Instruct is a checkpoint published by Moonshot AI, built on the DeepseekV3ForCausalLM architecture. SGLang 0.5.19, the newest indexed SGLang release, documents support. Official recipes or cookbooks cover vLLM, without pinning a stable release. TensorRT-LLM lists its architecture, but not this exact checkpoint.
SGLang documents support in 13 of 13 indexed releases, from 0.5.10 to 0.5.19.
- Documented hardware
- NVIDIA / CUDA and AMD / ROCm
- Documented quantization
- FP8
- Documented features
- Pipeline parallelism, Tensor parallelism, and Tool calling
- Configured context
- 131,072 tokens
- License
- modified-mit
Hardware, quantization and features count only statements about this checkpoint; engine-wide capabilities are listed with each release.
Hugging Face model card
- Publisher
- Moonshot AI
- Task
- text generation
- License
- modified-mit
- Library
- transformers
- Architecture
- DeepseekV3ForCausalLM
- Weight tensor elements
- 1T
- Tensor types
- F32, BF16, F8_E4M3
- Quantization
- fp8
- Configured context
- 131,072 tokens
- Weight access
- Public
- Repository created
- 2025-07-11
Model notes and sources
MoE checkpoint. Loading, tool parsing, and cache-transfer support are separate claims.
- SGLang v0.5.19 · docs/docs/supported-models/generative_models.mdx · 3 indexed revisions, v0.5.17 to v0.5.19
- moonshotai/Kimi-K2-Instruct publisher metadata
- SGLang v0.5.16 · docs_new/docs/supported-models/generative_models.mdx · 8 indexed revisions, v0.5.11 to v0.5.16
- SGLang v0.5.15.post1 · docs/supported_models/text_generation/generative_models.md · 9 indexed revisions, v0.5.10 to v0.5.15.post1
- moonshotai/Kimi-K2-Instruct · publisher model card
- moonshotai/Kimi-K2-Instruct · Hugging Face metadata
- moonshotai/Kimi-K2-Instruct · publisher configuration
- vllm · models/moonshotai/Kimi-K2-Instruct.yaml
- sglang · docs/cookbook/autoregressive/Moonshotai/Kimi-K2.mdx
- vllm · moonshotai/Kimi-K2.md
- Reviewed
Engine support
3 engines
| Engine | Release | Checkpoint support | Evidence scope | Last known supported |
|---|---|---|---|---|
| vLLM | Version not pinnedRecipe / cookbook | Documented recipeSource: vllm · models/moonshotai/Kimi-K2-Instruct.yamlReviewed | No supported release indexed | |
Features, dependencies and evidence for vLLM 0.28.0 | ||||
| SGLang | 0.5.19Newest indexed | Checkpoint evidenceSource: SGLang v0.5.19 · docs/docs/supported-models/generative_models.mdxReviewed | 0.5.19This release | |
Features, dependencies and evidence for SGLang 0.5.19 | ||||
| TensorRT-LLM | 1.2.1Newest indexed | Architecture listed; exact checkpoint unknown | No supported release indexed | |
Features, dependencies and evidence for TensorRT-LLM 1.2.1 | ||||