RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can SmolVLM Instruct run on the H200?

Yes. At 8K context the best fit is F16, using about 5.4 GB of the 134 GB usable on the H200. Estimated speed ceiling: ~1324 tokens/sec.

Memory by quant and context

Quant4K ctx8K ctxMax tok/s (est., 8K)
2-bit (Q2_K)2.0 GB Fits2.8 GB Fits5698
3-bit (Q3_K_M)2.3 GB Fits3.0 GB Fits4372
4-bit (Q4_K_M) •2.3 GB Fits3.0 GB Fits4316
5-bit (Q5_K_M)2.7 GB Fits3.5 GB Fits2999
6-bit (Q6_K)3.0 GB Fits3.7 GB Fits2606
8-bit (Q8_0) •3.0 GB Fits3.8 GB Fits2490
16-bit (F16/BF16) •4.6 GB Fits5.4 GB Fits1324

• real GGUF file size from ggml-org/SmolVLM-Instruct-GGUF. Speed is a bandwidth ceiling (4800 GB/s), not a benchmark. Methodology.

More

SmolVLM 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 (HuggingFaceTB/SmolVLM-Instruct), NVIDIA specs. Data as of 2026-11-17.