With 80 GB of VRAM, the H100 PCIe runs Gemma 3 4B IT at Q4_K_M entirely on the GPU with an 8K context (about 3.9 GB needed). At Q8_0 it fits entirely.
Q4_K_M · 8K context
The model needs about 3.9 GB and fits in the 77.6 GB available, with 73.7 GB to spare.
~3.9 GB (3.5 GB–4.4 GB) needed · estimate
Fits entirely in VRAM · 73.7 GB left
EstimateData updated 24 days agoSource: Run LLM Locally engine, Hugging Face config.json + model card
What makes this estimate less reliable
| Gemma 3 4B IT Q8_08K~5.4 GB · 208–351 tok/s estimate | Excellent |
|---|---|
| Gemma 3 4B IT Q6_K8K~4.6 GB · 238–421 tok/s estimate | Excellent |
| Gemma 3 4B IT Q5_K_M8K~4.2 GB · 252–456 tok/s estimate | Excellent |
| Gemma 3 4B IT Q4_K_M8K~3.9 GB · 266–495 tok/s estimate | Excellent |
| Gemma 3 4B IT IQ4_XS8K~3.7 GB · 277–525 tok/s estimate | Excellent |
| Gemma 3 4B IT Q3_K_M8K~3.5 GB · 286–549 tok/s estimate | Excellent |
Prices and links follow the market you choose.
NVIDIA
ExcellentRuns Gemma 3 4B IT Q4_K_M entirely in memory
AMD
ExcellentRuns Gemma 3 4B IT Q4_K_M entirely in memory
Unified memory PC
ExcellentRuns Gemma 3 4B IT Q4_K_M entirely in memory
No current offer imported for this market: the link opens a marketplace search.
