Delivery Exception Explanation
Given carrier scans, promised date, weather events, and exception codes, explain the current delay and next expected milestone without inventing an arrival time.
Compare 2 router rates · observed 9/4/2026
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 | 49% | 100/100 | $17.64 |
| Qwen3 235B A22B | mid | 69% | 100/100 | $70.80 |
| DeepSeek V3 | mid | 85% | 100/100 | $25.88 |
| Mistral Large 2407 | mid | 100% | 100/100 | $72.85 |
| Arcee Trinity Large Thinking | mid | 67% | 100/100 | $91.95 |
| GPT-5.4 | frontier | 100% | 100/100 | $434 |
| Claude Opus 4.7 | frontier | 90% | 100/100 | $1971 |
| Gemini 3.1 Pro Preview | frontier | 95% | 100/100 | $1505 |
| Gemma 4 E4B IT | small | 69% | 100/100 | $2.08 |
| Granite 4.1 8B | small | 74% | 100/100 | $5.38 |
| Ministral 8B Instruct 2410 | small | 63% | 100/100 | $9.84 |
| Qwen3 4B Instruct 2507 | small | 47% | 100/100 | $20.27 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-09T10:08:03.100Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock