RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can DeepSeek V3.1 run on the H200?

No. Even at 2-bit with 8K context, DeepSeek V3.1 needs about 230 GB, more than the 134 GB usable on the H200.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K) •230 GB No230 GB No232 GB No238 GB No—
3-bit (Q3_K_M) •299 GB No300 GB No301 GB No308 GB No—
4-bit (Q4_K_M) •378 GB No379 GB No380 GB No387 GB No—
5-bit (Q5_K_M) •444 GB No445 GB No446 GB No453 GB No—
6-bit (Q6_K) •514 GB No515 GB No516 GB No523 GB No—
8-bit (Q8_0) •665 GB No665 GB No667 GB No674 GB No—
16-bit (F16/BF16) •1251 GB No1251 GB No1253 GB No1259 GB No—

• real GGUF file size from unsloth/DeepSeek-V3.1-GGUF. Speed is a bandwidth ceiling (4800 GB/s), not a benchmark. Methodology.

Hardware that does fit it

Mac M3 Ultra 512GB

More

DeepSeek V3.1 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 (deepseek-ai/DeepSeek-V3.1), NVIDIA specs. Data as of 2026-11-17.