🔌
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 Codestral?
Yes — here's how fast and up to what context, with real measured numbers.
Codestral on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 320k | 78.9-94.1 t/s | ~0.7-0.8 (slow) |
| Q8 | yes | 404k | 157.8-188.1 t/s | ~1.3-1.6 (slow) |
| Q4 | yes | 442k | 281.9-335.9 t/s | ~2.3-2.8 (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 Codestral?▶
Yes. At Q4 it generates ~281.9-335.9 tok/s and fits up to 442k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~2.3-2.8 tok/s.
How much context fits?▶
At Q4, up to about 442k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.