🔌
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 OPT-2.7B (finetuned on WT2)?
Yes — here's how fast and up to what context, with real measured numbers.
OPT-2.7B (finetuned on WT2) on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1874k | 648.9-773.3 t/s | ~5.4-6.4 (slow) |
| Q8 | yes | 1915k | 1297.8-1546.7 t/s | ~10.8-12.9 (slow) |
| Q4 | yes | 1933k | 2317.5-2761.9 t/s | ~19.3-23 (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 OPT-2.7B (finetuned on WT2)?▶
Yes. At Q4 it generates ~2317.5-2761.9 tok/s and fits up to 1933k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~19.3-23 tok/s.
How much context fits?▶
At Q4, up to about 1933k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.