Can I run MiMo-V2.6-Distill-Qwen-9B on an NVIDIA H200?
Yes
Its published weights (18.8 GB) fit in the 127 GB an NVIDIA H200 can use, with room for ≈ 256K tokens of context.
Full analysis of MiMo-V2.6-Distill-Qwen-9B
Everything that runs on NVIDIA H200
Get the weights on Hugging Face
hf download XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9BFiles and context on NVIDIA H200
Each file of MiMo-V2.6-Distill-Qwen-9B (and of GGUF versions the site has analyzed), and how much context fits next to it, with the context cache in 16-bit or 8-bit.
| File | Size | On NVIDIA H200 | Context (16-bit) | Context (8-bit) |
|---|---|---|---|---|
| As published | 18.8 GB | Fits | 256K | 256K |
| BF16 | 17.9 GB | Fits | 256K | 256K |
| Q8_0 | 9.55 GB | Fits | 256K | 256K |
| Q6_K_L | 8.11 GB | Fits | 256K | 256K |
| Q6_K | 7.79 GB | Fits | 256K | 256K |
| Q6_K_S | 7.51 GB | Fits | 256K | 256K |
| Q5_K_M | 6.88 GB | Fits | 256K | 256K |
| Q5_K_S | 6.50 GB | Fits | 256K | 256K |
| Q4_K_L | 6.20 GB | Fits | 256K | 256K |
| Q4_1 | 5.94 GB | Fits | 256K | 256K |
| Q4_K_M | 5.84 GB | Fits | 256K | 256K |
| IQ4_NL | 5.83 GB | Fits | 256K | 256K |
| Q4_K_S | 5.48 GB | Fits | 256K | 256K |
| Q4_0 | 5.48 GB | Fits | 256K | 256K |
| IQ4_XS | 5.23 GB | Fits | 256K | 256K |
| IQ3_M | 4.85 GB | Fits | 256K | 256K |
| Q3_K_L | 4.66 GB | Fits | 256K | 256K |
| Q3_K_M | 4.48 GB | Fits | 256K | 256K |
| IQ3_XS | 4.27 GB | Fits | 256K | 256K |
| Q3_K_S | 4.26 GB | Fits | 256K | 256K |
| IQ3_XXS | 4.14 GB | Fits | 256K | 256K |
| Q2_K | 3.64 GB | Fits | 256K | 256K |
| IQ2_M | 3.54 GB | Fits | 256K | 256K |
| Q8_0 | 9.53 GB | Fits | 256K | 256K |