H100 NVL
10 records across 2 models (quants: FP8).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| H100 NVL | Qwen3 8B | FP8 | vLLM | – | 1202 | – | – | – | 2026-10-07 | community | – | source |
| H100 NVL | Qwen3 8B | FP8 | vLLM | – | 4160 | – | – | – | 2026-10-07 | community | – | source |
| H100 NVL | Qwen3 8B | FP8 | vLLM | – | 529 | – | – | – | 2026-10-07 | community | – | source |
| H100 NVL | Qwen3 8B | FP8 | vLLM | – | 634 | – | – | – | 2026-10-07 | community | – | source |
| H100 NVL | Qwen3.6-35B-A3B-FP8 | FP8 | vLLM | 1295.0 | – | – | – | tg512 | 2026-09-27 | community | – | source |
| H100 NVL | Qwen3.6-35B-A3B-FP8 | FP8 | vLLM | 1183.3 | – | – | – | tg512 | 2026-09-27 | community | – | source |
| H100 NVL | Qwen3.6-35B-A3B-FP8 | FP8 | vLLM | 362.3 | – | – | – | tg512 | 2026-09-27 | community | – | source |
| H100 NVL | Qwen3.6-35B-A3B-FP8 | FP8 | vLLM | 148.3 | – | – | – | tg512 | 2026-09-27 | community | – | source |
| H100 NVL | Qwen3.6-35B-A3B-FP8 | FP8 | vLLM | 138.8 | – | – | – | tg512 | 2026-09-27 | community | – | source |
| H100 NVL | Qwen3.6-35B-A3B-FP8 | FP8 | vLLM | 41.4 | – | – | – | tg512 | 2026-09-27 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085