Moonshot AI
Kimi K2.5
Kimi family, 1000B parameters, recommended as Q4_K_M for first-pass local usage in V1.
Params
1000B
Context
256K
License
Custom
Best runtime
vLLM
Recommended hardware
First-pass fit across priority GPUs
| Hardware | Fit | Decode | Safe ctx |
|---|---|---|---|
| NVIDIA A10 24GB | Too heavy | 2 tok/s | 4K |
| NVIDIA A100 40GB | Too heavy | 3.5 tok/s | 4K |
| NVIDIA A100 80GB | Too heavy | 4.6 tok/s | 4K |
| NVIDIA A16 64GB | Too heavy | 2 tok/s | 4K |