🔌
New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
// RUN — FITS
Can the NVIDIA H200 run Gemma 3 4B?
Yes — here's how fast and up to what context, with real measured numbers.
Gemma 3 4B on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1834k | 438-522 t/s | ~3.7-4.4 (slow) |
| Q8 | yes | 1895k | 876-1044 t/s | ~7.3-8.7 (slow) |
| Q4 | yes | 1922k | 1564.3-1864.3 t/s | ~13-15.5 (slow) |
Estimate (memory-bound). Beyond "max context" the KV cache spills to system RAM offload (depends on your system RAM) and speed drops to the "beyond" figure.
FAQ
Can the NVIDIA H200 run Gemma 3 4B?▶
Yes. At Q4 it generates ~1564.3-1864.3 tok/s and fits up to 1922k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~13-15.5 tok/s.
How much context fits?▶
At Q4, up to about 1922k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.