RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can deepseek coder 7b instruct v1.5 run on the H100 SXM?

Yes. At 8K context the best fit is F16, using about 17.2 GB of the 76.0 GB usable on the H100 SXM. Estimated speed ceiling: ~242 tokens/sec.

Memory by quant and context

Quant4K ctxMax tok/s (est., 8K)
2-bit (Q2_K) •5.0 GB Fits1232
3-bit (Q3_K_M) •5.7 GB Fits968
4-bit (Q4_K_M) •6.4 GB Fits793
5-bit (Q5_K_M) •7.1 GB Fits680
6-bit (Q6_K) •7.8 GB Fits590
8-bit (Q8_0)9.3 GB Fits456
16-bit (F16/BF16)15.3 GB Fits242

• real GGUF file size from mradermacher/deepseek-coder-7b-instruct-v1.5-i1-GGUF. Speed is a bandwidth ceiling (3350 GB/s), not a benchmark. Methodology.

More

deepseek coder 7b instruct v1.5 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 (deepseek-ai/deepseek-coder-7b-instruct-v1.5), NVIDIA specs. Data as of 2026-11-17.