Churn Save Agent
Diagnose cancellation intent and offer only policy-allowed remedies before completing or retaining the account.
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 | 34% | 100/100 | $30.31 |
| Qwen3 235B A22B | mid | 34% | 100/100 | $195 |
| DeepSeek V3 | mid | 46% | 100/100 | $72.75 |
| Mistral Large 2407 | mid | 76% | 100/100 | $155 |
| Arcee Trinity Large Thinking | mid | 40% | 100/100 | $100 |
| GPT-5.4 | frontier | 100% | 100/100 | $1459 |
| Claude Opus 4.7 | frontier | 94% | 100/100 | $4044 |
| Gemini 3.1 Pro Preview | frontier | 77% | 100/100 | $1758 |
| Gemma 4 E4B IT | small | 63% | 100/100 | $7.36 |
| Granite 4.1 8B | small | 33% | 100/100 | $8.81 |
| Ministral 8B Instruct 2410 | small | 29% | 100/100 | $14.05 |
| Qwen3 4B Instruct 2507 | small | 4% | 100/100 | $83.49 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-10T22:53:02.765Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock