8x NVIDIA H100 80GB HBM3
6 records across 1 models (quants: BF16).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 8x NVIDIA H100 80GB HBM3 | GLM-5.3-Flash | BF16 | vLLM | 754.1 | – | – | – | tg256 | 2026-09-12 | community | – | source |
| 8x NVIDIA H100 80GB HBM3 | GLM-5.3-Flash | BF16 | vLLM | 368.8 | – | – | – | tg256 | 2026-09-12 | community | – | source |
| 8x NVIDIA H100 80GB HBM3 | GLM-5.3-Flash | BF16 | vLLM | 249.3 | – | – | – | tg256 | 2026-09-12 | community | – | source |
| 8x NVIDIA H100 80GB HBM3 | GLM-5.3-Flash | BF16 | vLLM | 237.0 | – | – | – | tg256 | 2026-09-12 | community | – | source |
| 8x NVIDIA H100 80GB HBM3 | GLM-5.3-Flash | BF16 | vLLM | 229.8 | – | – | – | tg256 | 2026-09-12 | community | – | source |
| 8x NVIDIA H100 80GB HBM3 | GLM-5.3-Flash | BF16 | vLLM | 162.9 | – | – | – | tg256 | 2026-09-12 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085