Load-Test Breaking-Point Summary
Given time-series load-test results, identify saturation point, first failed SLA, likely bottleneck, and the evidence supporting each conclusion.
View token rates · observed 9/11/2026
Compare 2 router rates · observed 9/11/2026
Compare 2 router rates · observed 9/11/2026
All 12 model results and methodology
Generated 100/100 examples
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 2% | 100/100 | $36.05 |
| Qwen3 235B A22B | mid | 0% | 100/100 | $202 |
| DeepSeek V3 | mid | 27% | 100/100 | $68.99 |
| Mistral Large 2407 | mid | 14% | 100/100 | $189 |
| Arcee Trinity Large Thinking | mid | 24% | 100/100 | $119 |
| GPT-5.4 | frontier | 83% | 100/100 | $1330 |
| Claude Opus 4.7 | frontier | 32% | 100/100 | $3675 |
| Gemini 3.1 Pro Preview | frontier | 43% | 100/100 | $2551 |
| Gemma 4 E4B IT | small | 20% | 100/100 | $6.81 |
| Granite 4.1 8B | small | 0% | 100/100 | $10.99 |
| Ministral 8B Instruct 2410 | small | 2% | 100/100 | $26.97 |
| Qwen3 4B Instruct 2507 | small | 0% | 100/100 | $67.50 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-09T03:07:02.611Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock