8xH100
8 records across 1 models (quants: FP8).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 8xH100 | GLM-5.3-Flash | FP8 | vLLM | 25101 | – | – | – | 65536 | 2026-10-06 | sourced | – | source |
| 8xH100 | GLM-5.3-Flash | FP8 | vLLM | 24263 | – | – | – | 65536 | 2026-10-06 | sourced | – | source |
| 8xH100 | GLM-5.3-Flash | FP8 | vLLM | 23357 | – | – | – | 65536 | 2026-10-06 | sourced | – | source |
| 8xH100 | GLM-5.3-Flash | FP8 | vLLM | 23141 | – | – | – | 65536 | 2026-10-06 | sourced | – | source |
| 8xH100 | GLM-5.3-Flash | FP8 | vLLM | 21522 | – | – | – | 8192 | 2026-10-06 | sourced | – | source |
| 8xH100 | GLM-5.3-Flash | FP8 | vLLM | 20189 | – | – | – | 8192 | 2026-10-06 | sourced | – | source |
| 8xH100 | GLM-5.3-Flash | FP8 | vLLM | 16717 | – | – | – | 8192 | 2026-10-06 | sourced | – | source |
| 8xH100 | GLM-5.3-Flash | FP8 | vLLM | 16424 | – | – | – | 8192 | 2026-10-06 | sourced | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085