PatRe - Office Action Decision (OA-RO): leaderboard

PatRe (2026) models patent examination on 480 recent USPTO cases across all eight IPC sections. Acting as the examiner, the model writes an office action from the claims together with the references the examiner and applicant cited (reference oracle). The score is the percent of office actions whose decision (allowance, non-final or final rejection, Ex parte Quayle) matches the examiner; higher is better.

Metric: Decision accuracy (%): the generated office action's decision (allowance, non-final or final rejection, Ex parte Quayle) matches the examiner's, office action generation under reference oracle: the model also receives the references the examiner and applicant cited and must pick the relevant ones, over PatRe (480 recent USPTO patent examination histories across all eight IPC sections, with office actions, rebuttals, claim versions and cited references), temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around September 2028. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5 Mini50#373
2DeepSeek V3.249.7#425
3Qwen 3.5 27B47.6#380
4Gemini 2.5 Flash46.4#466
5Llama 3.1 8B Instruct45.6#4655
6Gemma 3 27B (IT)45.4#1341
7Qwen 3.5 9B44.3#838
8Gemma 3 12B (IT)43.4#1907
9GPT-4o Mini43.3#1680
10Llama 3.3 70B Instruct9.7#1394

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=patre-office-action-decision-oa-ro · How It Works · Data refreshed daily, snapshot 2026-11-22.