🔌
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 Qwen-VLA?
Yes — here's how fast and up to what context, with real measured numbers.
Qwen-VLA on the NVIDIA B200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1258k | 567-675.7 t/s | ~2.8-3.4 (slow) |
| Q8 | yes | 1297k | 1134-1351.5 t/s | ~5.7-6.8 (slow) |
| Q4 | yes | 1314k | 2025-2413.3 t/s | ~10.1-12.1 (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 Qwen-VLA?▶
Yes. At Q4 it generates ~2025-2413.3 tok/s and fits up to 1314k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~10.1-12.1 tok/s.
How much context fits?▶
At Q4, up to about 1314k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.