Loading…
Loading…
Models / Kimi-K3
1 model (~2.78e+03B) · ↓ 1M downloads a month · from $132.05/h on 8x Nvidia B300 right now. Each model's versions do the same job; quantized ones (AWQ, FP8, 4-bit) need less GPU memory and run on cheaper machines.
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originalmoonshotai/Kimi-K3 | vLLM | 1561 GB | 1760.6 GB | $132.05/h 8x Nvidia B300 |
| NVFP4 (nvidia)nvidia/Kimi-K3-NVFP4 | vLLM | 1610 GB | 1815.1 GB | $132.05/h 8x Nvidia B300 |
| Main (unsloth)unsloth/Kimi-K3 | vLLM | 1561 GB | 1760.6 GB | $132.05/h 8x Nvidia B300 |
Model details come from Hugging Face and are refreshed daily; prices follow the live offers. GPU memory is GridPup's estimate: the weights, room for an 8,192-token conversation (chat models) and the runtime's own memory. Large models are split across the GPUs of one machine ("4x RTX 4090").