🔌
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_base?
Yes — here's how fast and up to what context, with real measured numbers.
Ankh_base on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1934k | 2367.6-2821.6 t/s | ~19.7-23.5 (slow) |
| Q8 | yes | 1945k | 4735.1-5643.2 t/s | ~39.5-47 (slow) |
| Q4 | yes | 1950k | 8455.6-10077.2 t/s | ~70.5-84 (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_base?▶
Yes. At Q4 it generates ~8455.6-10077.2 tok/s and fits up to 1950k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~70.5-84 tok/s.
How much context fits?▶
At Q4, up to about 1950k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.