🔌
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 Japanese-LM-3.6B?
Yes — here's how fast and up to what context, with real measured numbers.
Japanese-LM-3.6B on the NVIDIA B200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 2563k | 811.1-966.7 t/s | ~4.1-4.8 (slow) |
| Q8 | yes | 2617k | 1622.2-1933.3 t/s | ~8.1-9.7 (slow) |
| Q4 | yes | 2642k | 2896.8-3452.4 t/s | ~14.5-17.3 (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 Japanese-LM-3.6B?▶
Yes. At Q4 it generates ~2896.8-3452.4 tok/s and fits up to 2642k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~14.5-17.3 tok/s.
How much context fits?▶
At Q4, up to about 2642k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.