RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can Qwen3.8 27B run on the H200?

Yes. At 8K context the best fit is F16, using about 52.3 GB of the 134 GB usable on the H200. Estimated speed ceiling: ~88 tokens/sec.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K)10.8 GB Fits11.1 GB Fits12.6 GB Fits18.6 GB Fits461
3-bit (Q3_K_M)13.8 GB Fits14.0 GB Fits15.5 GB Fits21.5 GB Fits354
4-bit (Q4_K_M) •16.6 GB Fits16.8 GB Fits18.3 GB Fits24.3 GB Fits290
5-bit (Q5_K_M) •19.6 GB Fits19.9 GB Fits21.4 GB Fits27.4 GB Fits242
6-bit (Q6_K) •21.6 GB Fits21.8 GB Fits23.3 GB Fits29.3 GB Fits218
8-bit (Q8_0) •28.2 GB Fits28.4 GB Fits29.9 GB Fits35.9 GB Fits165
16-bit (F16/BF16) •52.0 GB Fits52.3 GB Fits53.8 GB Fits59.8 GB Fits88

• real GGUF file size from unsloth/Qwen3.8-27B-GGUF. Speed is a bandwidth ceiling (4800 GB/s), not a benchmark. Methodology.

More

Qwen3.8 27B 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 (Qwen/Qwen3.8-27B), NVIDIA specs. Data as of 2026-11-17.