Can Mistral 7B Instruct v0.1 run on the H200?
Yes. At 8K context the best fit is F16, using about 14.9 GB of the 134 GB usable on the H200. Estimated speed ceiling: ~331 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|
| 2-bit (Q2_K) • | 3.8 GB Fits | 4.3 GB Fits | 7.3 GB Fits | 1557 |
| 3-bit (Q3_K_M) • | 4.2 GB Fits | 4.7 GB Fits | 7.7 GB Fits | 1364 |
| 4-bit (Q4_K_M) • | 5.0 GB Fits | 5.5 GB Fits | 8.5 GB Fits | 1099 |
| 5-bit (Q5_K_M) • | 5.7 GB Fits | 6.2 GB Fits | 9.2 GB Fits | 935 |
| 6-bit (Q6_K) • | 6.5 GB Fits | 7.0 GB Fits | 10.0 GB Fits | 808 |
| 8-bit (Q8_0) • | 8.1 GB Fits | 8.6 GB Fits | 11.6 GB Fits | 624 |
| 16-bit (F16/BF16) | 14.4 GB Fits | 14.9 GB Fits | 17.9 GB Fits | 331 |
• real GGUF file size from TheBloke/Mistral-7B-Instruct-v0.1-GGUF. Speed is a bandwidth ceiling (4800 GB/s), not a benchmark. Methodology.
More
Mistral 7B Instruct v0.1 on every GPU and Mac · Every model that runs on the H200 · Other models that fit 64 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (mistralai/Mistral-7B-Instruct-v0.1), NVIDIA specs. Data as of 2026-11-17.