Can Qwen2.5 Coder 14B Instruct run on the H100 SXM?
Yes. At 8K context the best fit is F16, using about 29.7 GB of the 76.0 GB usable on the H100 SXM. Estimated speed ceiling: ~113 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|
| 2-bit (Q2_K) | 6.6 GB Fits | 7.3 GB Fits | 11.8 GB Fits | 605 |
| 3-bit (Q3_K_M) | 8.2 GB Fits | 8.9 GB Fits | 13.4 GB Fits | 464 |
| 4-bit (Q4_K_M) | 9.8 GB Fits | 10.6 GB Fits | 15.1 GB Fits | 371 |
| 5-bit (Q5_K_M) | 11.2 GB Fits | 12.0 GB Fits | 16.5 GB Fits | 318 |
| 6-bit (Q6_K) | 12.7 GB Fits | 13.5 GB Fits | 18.0 GB Fits | 277 |
| 8-bit (Q8_0) | 16.1 GB Fits | 16.8 GB Fits | 21.3 GB Fits | 213 |
| 16-bit (F16/BF16) | 29.0 GB Fits | 29.7 GB Fits | 34.2 GB Fits | 113 |
• real GGUF file size. Speed is a bandwidth ceiling (3350 GB/s), not a benchmark. Methodology.
More
Qwen2.5 Coder 14B Instruct on every GPU and Mac · Every model that runs on the H100 SXM · Other models that fit 64 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (Qwen/Qwen2.5-Coder-14B-Instruct), NVIDIA specs. Data as of 2026-11-17.