RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can Phi 3.5 vision instruct run on the H200?

Yes. At 8K context the best fit is F16, using about 11.2 GB of the 134 GB usable on the H200. Estimated speed ceiling: ~579 tokens/sec.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K)3.4 GB Fits4.9 GB Fits13.9 GB Fits49.9 GB Fits3087
3-bit (Q3_K_M)3.8 GB Fits5.3 GB Fits14.3 GB Fits50.3 GB Fits2368
4-bit (Q4_K_M)4.3 GB Fits5.8 GB Fits14.8 GB Fits50.8 GB Fits1894
5-bit (Q5_K_M)4.7 GB Fits6.2 GB Fits15.2 GB Fits51.2 GB Fits1625
6-bit (Q6_K)5.1 GB Fits6.6 GB Fits15.6 GB Fits51.6 GB Fits1412
8-bit (Q8_0)6.1 GB Fits7.6 GB Fits16.6 GB Fits52.6 GB Fits1089
16-bit (F16/BF16)9.7 GB Fits11.2 GB Fits20.2 GB Fits56.2 GB Fits579

• real GGUF file size. Speed is a bandwidth ceiling (4800 GB/s), not a benchmark. Methodology.

More

Phi 3.5 vision instruct 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 (microsoft/Phi-3.5-vision-instruct), NVIDIA specs. Data as of 2026-11-17.