RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?OllamaStart hereAPI pricesVRAM calculatorNew

Can Gemma 3 4B PT run on the H100 SXM?

Yes. At 8K context the best fit is F16, using about 8.4 GB of the 76.0 GB usable on the H100 SXM. Estimated speed ceiling: ~431 tokens/sec.

+ 0.5 GB with image input. The figures on this page are for text only. With image input enabled, the vision projector (mmproj) file loads alongside the model: F16 at 8K context then needs about 9.0 GB (still fits). llama.cpp's -hf downloads and loads it automatically unless you add --no-mmproj. Calculator with image input on.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K) •2.7 GB Fits2.8 GB Fits3.2 GB Fits5.1 GB Fits1938
3-bit (Q3_K_M) •3.0 GB Fits3.1 GB Fits3.6 GB Fits5.5 GB Fits1597
4-bit (Q4_K_M) •3.4 GB Fits3.5 GB Fits4.0 GB Fits5.8 GB Fits1346
5-bit (Q5_K_M) •3.7 GB Fits3.8 GB Fits4.3 GB Fits6.1 GB Fits1184
6-bit (Q6_K) •4.1 GB Fits4.1 GB Fits4.6 GB Fits6.5 GB Fits1050
8-bit (Q8_0) •4.9 GB Fits5.0 GB Fits5.5 GB Fits7.4 GB Fits811
16-bit (F16/BF16) •8.3 GB Fits8.4 GB Fits8.9 GB Fits10.7 GB Fits431

• real GGUF file size from mradermacher/gemma-3-4b-pt-GGUF. Speed is a bandwidth ceiling (3350 GB/s), not a benchmark. Methodology.

What fits instead

"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.

More

Gemma 3 4B PT 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

Guides: Best GPU for local LLMs by budget · Can I run gpt-oss, Llama 70B or Qwen3? · Mac vs PC for local LLMs

Data: Hugging Face Hub (google/gemma-3-4b-pt), NVIDIA specs. Model data last checked 2026-11-18; refreshed hourly.