Vendor Entity Deduplication
Given vendor-master records, identify likely duplicates across name, address, tax ID, and payment details while distinguishing related but separate legal entities.
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 | 19% | 100/100 | $51.71 |
| Qwen3 235B A22B | mid | 56% | 100/100 | $398 |
| DeepSeek V3 | mid | 73% | 100/100 | $145 |
| Mistral Large 2407 | mid | 53% | 100/100 | $337 |
| Arcee Trinity Large Thinking | mid | 57% | 100/100 | $169 |
| GPT-5.4 | frontier | 77% | 100/100 | $3084 |
| Claude Opus 4.7 | frontier | 34% | 100/100 | $5787 |
| Gemini 3.1 Pro Preview | frontier | 1% | 99/100 | $2627 |
| Gemma 4 E4B IT | small | 71% | 100/100 | $18.43 |
| Granite 4.1 8B | small | 36% | 100/100 | $19.47 |
| Ministral 8B Instruct 2410 | small | 11% | 100/100 | $43.45 |
| Qwen3 4B Instruct 2507 | small | 29% | 100/100 | $140 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-10T08:59:56.429Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock