Information Retrieval on Patents from USPTO office actions
91.5RecallOpus-4.6
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Opus-4.6Category=Frontier models2026.06 | 91.5 | 91.5 | 93.9 | |
| GPT-5.4Category=Frontier models2026.06 | 90.9 | 90.9 | 91.9 | |
| Harness-1Parameters=20B, Category=Open-source small models2026.06 | 89 | 89 | 97.3 | |
| Context-1Parameters=20B, Category=Open-source small models2026.06 | 85.7 | 85.7 | 97.5 | |
| Kimi-K2.5Category=Frontier models2026.06 | 83.3 | 83.3 | 94.5 | |
| Sonnet-4.6Category=Frontier models2026.06 | 83.2 | 83.2 | 84.5 | |
| Tongyi DeepResearchParameters=30B, Category=Open-source small models2026.06 | 78 | 78 | 78 | |
| GPT-OSSParameters=120B, Category=Frontier models2026.06 | 57 | 57 | 82.4 | |
| Search-R1Parameters=32B, Category=Open-source small models2026.06 | 40.5 | 40.5 | 40.5 | |
| GPT-OSSParameters=20B, Category=Open-source small models2026.06 | 39.3 | 39.3 | 74.2 | |
| Qwen3Parameters=32B, Category=Open-source small models2026.06 | 25.7 | 25.7 | 57.7 |