Vendor Onboarding Agent
Validate identity, tax, banking, sanctions, and required documents; create or route the vendor record.
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 99/100 examples — partial set accepted at the success threshold
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 69% | 99/99 | $32.14 |
| Qwen3 235B A22B | mid | 71% | 99/99 | $233 |
| DeepSeek V3 | mid | 74% | 99/99 | $80.35 |
| Mistral Large 2407 | mid | 80% | 99/99 | $195 |
| Arcee Trinity Large Thinking | mid | 70% | 99/99 | $121 |
| GPT-5.4 | frontier | 97% | 99/99 | $2151 |
| Claude Opus 4.7 | frontier | 91% | 99/99 | $4788 |
| Gemini 3.1 Pro Preview | frontier | 59% | 99/99 | $2444 |
| Gemma 4 E4B IT | small | 76% | 99/99 | $9.96 |
| Granite 4.1 8B | small | 60% | 99/99 | $10.89 |
| Ministral 8B Instruct 2410 | small | 44% | 99/99 | $21.95 |
| Qwen3 4B Instruct 2507 | small | 32% | 99/99 | $103 |
LLM-judge pass rate on 99 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-11T02:17:02.592Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock