🔌
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 WizardCoder-15.5B?
Yes — here's how fast and up to what context, with real measured numbers.
WizardCoder-15.5B on the NVIDIA H200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 371k | 113-134.7 t/s | ~0.9-1.1 (slow) |
| Q8 | yes | 430k | 226.1-269.4 t/s | ~1.9-2.2 (slow) |
| Q4 | yes | 456k | 403.7-481.1 t/s | ~3.4-4 (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 WizardCoder-15.5B?▶
Yes. At Q4 it generates ~403.7-481.1 tok/s and fits up to 456k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~3.4-4 tok/s.
How much context fits?▶
At Q4, up to about 456k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.