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

Can Granite 3.3 8B Instruct run on the H100 SXM?

Yes. At 8K context the best fit is F16, using about 17.0 GB of the 76.0 GB usable on the H100 SXM. Estimated speed ceiling: ~205 tokens/sec.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K) •4.0 GB Fits4.6 GB Fits8.4 GB Fits23.4 GB Fits1079
3-bit (Q3_K_M) •4.8 GB Fits5.5 GB Fits9.2 GB Fits24.2 GB Fits838
4-bit (Q4_K_M) •5.7 GB Fits6.3 GB Fits10.1 GB Fits25.1 GB Fits678
5-bit (Q5_K_M) •6.5 GB Fits7.1 GB Fits10.9 GB Fits25.9 GB Fits578
6-bit (Q6_K) •7.4 GB Fits8.0 GB Fits11.7 GB Fits26.7 GB Fits500
8-bit (Q8_0) •9.2 GB Fits9.8 GB Fits13.6 GB Fits28.6 GB Fits386
16-bit (F16/BF16) •16.3 GB Fits17.0 GB Fits20.7 GB Fits35.7 GB Fits205

• real GGUF file size from ibm-granite/granite-3.3-8b-instruct-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

Granite 3.3 8B Instruct 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 (ibm-granite/granite-3.3-8b-instruct), NVIDIA specs. Model data last checked 2026-11-17; refreshed hourly.