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 GPT-2 Medium (FlashAttention)?
Yes — here's how fast and up to what context, with real measured numbers.
GPT-2 Medium (FlashAttention) on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1946k | 4935.2-5881.7 t/s | ~41.1-49 (slow) |
| Q8 | yes | 1951k | 9870.4-11763.4 t/s | ~82.3-98 (slow) |
| Q4 | yes | 1953k | 17625.8-21006 t/s | ~146.9-175.1 (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 GPT-2 Medium (FlashAttention)?▶
Yes. At Q4 it generates ~17625.8-21006 tok/s and fits up to 1953k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~146.9-175.1 tok/s.
How much context fits?▶
At Q4, up to about 1953k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.