🔌
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 B200 run LLaMA-7B?
Yes — here's how fast and up to what context, with real measured numbers.
LLaMA-7B on the NVIDIA B200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1234k | 435.8-519.4 t/s | ~2.2-2.6 (slow) |
| Q8 | yes | 1285k | 871.6-1038.8 t/s | ~4.4-5.2 (slow) |
| Q4 | yes | 1308k | 1556.5-1855 t/s | ~7.8-9.3 (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 B200 run LLaMA-7B?▶
Yes. At Q4 it generates ~1556.5-1855 tok/s and fits up to 1308k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~7.8-9.3 tok/s.
How much context fits?▶
At Q4, up to about 1308k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.