4x NVIDIA GB300
4 records across 1 models (quants: unstated).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 4x NVIDIA GB300 | DeepSeek-V3.1 | unstated | vLLM | 1168.47 | – | – | 4434.9 | – | 2026-10 | sourced | – | source |
| 4x NVIDIA GB300 | DeepSeek-V3.1 | unstated | vLLM | 982.28 | – | – | 5609.71 | – | 2026-10 | sourced | – | source |
| 4x NVIDIA GB300 | DeepSeek-V3.1 | unstated | vLLM | – | 10575.95 | – | 50052.37 | – | 2026-10 | sourced | – | source |
| 4x NVIDIA GB300 | DeepSeek-V3.1 | unstated | vLLM | – | 3083.34 | – | 171527.46 | – | 2026-10 | sourced | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085