🔌
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 KoGPT?
Yes — here's how fast and up to what context, with real measured numbers.
KoGPT on the NVIDIA B200
| Quant | Fits? | Max context | Speed | Beyond context |
|---|---|---|---|---|
| FP16 | yes | 1242k | 473.5-564.3 t/s | ~2.4-2.8 (slow) |
| Q8 | yes | 1289k | 947.1-1128.7 t/s | ~4.7-5.6 (slow) |
| Q4 | yes | 1310k | 1691.2-2015.5 t/s | ~8.5-10.1 (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 KoGPT?▶
Yes. At Q4 it generates ~1691.2-2015.5 tok/s and fits up to 1310k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~8.5-10.1 tok/s.
How much context fits?▶
At Q4, up to about 1310k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.