🔌
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 Ankh_large?
Yes — here's how fast and up to what context, with real measured numbers.
Ankh_large on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1898k | 922.1-1098.9 t/s | ~7.7-9.2 (slow) |
| Q8 | yes | 1927k | 1844.2-2197.9 t/s | ~15.4-18.3 (slow) |
| Q4 | yes | 1940k | 3293.2-3924.8 t/s | ~27.4-32.7 (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 Ankh_large?▶
Yes. At Q4 it generates ~3293.2-3924.8 tok/s and fits up to 1940k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~27.4-32.7 tok/s.
How much context fits?▶
At Q4, up to about 1940k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.