AGORA - Law: leaderboard

Metric: Accuracy (%) on the 35 law questions (285 documents), archive-grounded agentic document reasoning with verifiable numeric answers over the domain collection of Markdown workplace documents, agent in the mini-swe-agent harness (bash tool only), temperature 1.0 with reasoning effort at the maximum each model supports; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 8 models tracked.

Top models

#ModelScore
1GPT-5.5 (xHigh)62.86
2GLM-5.151.43
3Gemini 3.1 Pro (Preview) (High)51.43
4DeepSeek V4 Pro (Max)51.43
5DeepSeek V4 Flash (Max)31.43
6Qwen 3.5 35B A3B17.14
7Qwen 3.5 9B5.71
8Gemini 3.1 Flash Lite (High)2.86

Interactive version: theaggregate.ai/benchmark?slug=agora-law · How It Works · Data refreshed daily, snapshot 2026-09-29.