Models / Kimi K2generated from capability-matrix.json
Kimi K2
Moonshot's Kimi K2 line, K2.7-Code included. Runs on DeepSeek-V3's code path.
chatcodeexperts: sparse +sharedsafetensors, GGUF
Get it
No vetted checkpoint yet. You can pull any GGUF of this family by repo and quant. goinfer checks it fits before downloading, but nobody here has timed it.
goinfer-chat pull <owner>/<repo>:<quant>What hasn't been shown
- No check of its own. It runs DeepSeek-V3's code, and DeepSeek-V3's numbers stand in for it.
- No vetted checkpoint, so there's no fit verdict, no speed and no tool-calling result for this family.
How sure we are
Shares another family's checkSame forward code as DeepSeek-V3, so it has no numbers of its own.
shared-path: deepseek_v3 · what parity-gated means
Measured speed · decode, tokens per second
not measured — no vetted checkpoint to measure
Architecture, for the curious
- model_type
- kimi_k2
- design
- latent-KV (MLA)
- experts
- sparse +shared
- attention window
- none
- QK-norm
- no
- RoPE
- partial
- norm
- RMSNorm, pre-norm
- activation
- SwiGLU
- tied head
- no
- modality
- text
- GPU-resident
- eligible
registry description: Moonshot Kimi K2 / K2.5 / K2.6 / K2.7-Code (DeepseekV3 arch: MLA + DeepSeekMoE — same arch across the K2.x line)