Runs
All benchmark runs
One run = one model × precision on one hardware target. Open a run for every concurrency level and accuracy task.
| Model | Hardware | Engine | Peak users (SLO) | Date | Run ID |
|---|---|---|---|---|---|
| Qwen3-8B · BF16 | 1× H100-80GB | vLLM 0.30.0 | ≥ 512 | 2026-09-24 | 20260924T125045Z_h100-80gbx1_qwen3-8b_bf16 |
| Qwen3-8B · FP8 | 1× H100-80GB | vLLM 0.30.0 | ≥ 512 | 2026-09-24 | 20260924T225750Z_h100-80gbx1_qwen3-8b_fp8 |