🔌
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 Hymba?
Yes — here's how fast and up to what context, with real measured numbers.
Hymba on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1911k | 1168-1392 t/s | ~9.7-11.6 (slow) |
| Q8 | yes | 1934k | 2336-2784 t/s | ~19.5-23.2 (slow) |
| Q4 | yes | 1944k | 4171.4-4971.4 t/s | ~34.8-41.4 (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 Hymba?▶
Yes. At Q4 it generates ~4171.4-4971.4 tok/s and fits up to 1944k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~34.8-41.4 tok/s.
How much context fits?▶
At Q4, up to about 1944k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.