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 DeBERTaV3large?
Yes — here's how fast and up to what context, with real measured numbers.
DeBERTaV3large on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1944k | 4191.4-4995.2 t/s | ~34.9-41.6 (slow) |
| Q8 | yes | 1950k | 8382.8-9990.4 t/s | ~69.9-83.3 (slow) |
| Q4 | yes | 1953k | 14969.2-17840.1 t/s | ~124.7-148.7 (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 DeBERTaV3large?▶
Yes. At Q4 it generates ~14969.2-17840.1 tok/s and fits up to 1953k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~124.7-148.7 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.