Patent Prior-Art Plain-Language Summary
Given a patent claim and several candidate prior-art passages, summarize the claimed novelty and the most relevant overlaps without making a legal conclusion.
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 | 54% | 100/100 | $22.95 |
| Qwen3 235B A22B | mid | 24% | 100/100 | $132 |
| DeepSeek V3 | mid | 100% | 100/100 | $47.23 |
| Mistral Large 2407 | mid | 97% | 100/100 | $116 |
| Arcee Trinity Large Thinking | mid | 76% | 100/100 | $88.73 |
| GPT-5.4 | frontier | 100% | 100/100 | $1352 |
| Claude Opus 4.7 | frontier | 95% | 100/100 | $3532 |
| Gemini 3.1 Pro Preview | frontier | 98% | 100/100 | $1905 |
| Gemma 4 E4B IT | small | 99% | 100/100 | $3.71 |
| Granite 4.1 8B | small | 45% | 100/100 | $8.11 |
| Ministral 8B Instruct 2410 | small | 73% | 100/100 | $14.87 |
| Qwen3 4B Instruct 2507 | small | 37% | 100/100 | $40.73 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-08T19:41:02.637Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock