Product Attribute Tagging
Given a product title, description, and allowed taxonomy, assign supported color, material, style, occasion, and feature tags without inventing attributes.
View token rates
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 | 44% | 100/100 | $17.38 |
| Qwen3 235B A22B | mid | 36% | 100/100 | $24.34 |
| DeepSeek V3 | mid | 57% | 100/100 | $15.01 |
| Mistral Large 2407 | mid | 59% | 100/100 | $31.70 |
| Arcee Trinity Large Thinking | mid | 56% | 100/100 | $88.33 |
| GPT-5.4 | frontier | 74% | 100/100 | $169 |
| Claude Opus 4.7 | frontier | 62% | 100/100 | $478 |
| Gemini 3.1 Pro Preview | frontier | 61% | 100/100 | $958 |
| Gemma 4 E4B IT | small | 36% | 100/100 | $1.67 |
| Granite 4.1 8B | small | 40% | 100/100 | $2.06 |
| Ministral 8B Instruct 2410 | small | 33% | 100/100 | $5.73 |
| Qwen3 4B Instruct 2507 | small | 46% | 100/100 | $8.14 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-09T09:53:02.599Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock