Baseline keys
| Baseline key | Backend | Status | Regression | Quality | Cost / 1M tokens | Last run |
|---|---|---|---|---|---|---|
| llama8b_vllm_h100 | vllm | ok | inconclusive | 90% | $0.55 +1.0% | 2026-10-30T03:10:40.349700+00:00 |
| qwen7b_fp8_h100 | vllm | failed | regressed | 92% | $0.42 -18.7% | 2026-10-30T03:17:40.349700+00:00 |
| qwen7b_vllm_h100 | vllm | ok | held | 90% | $0.55 +3.8% | 2026-10-30T03:15:40.349700+00:00 |