New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
Can the NVIDIA H200 run GemNet-T (OC20)?
Yes — here's how fast and up to what context, with real measured numbers.
GemNet-T (OC20) on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1956k | 922105.3-1098947.4 t/s | ~7684.2-9157.9 (slow) |
| Q8 | yes | 1956k | 1844210.5-2197894.7 t/s | ~15368.4-18315.8 (slow) |
| Q4 | yes | 1956k | 3293233.1-3924812 t/s | ~27443.6-32706.8 (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 GemNet-T (OC20)?▶
Yes. At Q4 it generates ~3293233.1-3924812 tok/s and fits up to 1956k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~27443.6-32706.8 tok/s.
How much context fits?▶
At Q4, up to about 1956k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.