ORAgentBench - Medium: leaderboard

Metric: Pass rate (%) on the 41 medium tasks: a task passes when the submission is feasible and its normalized quality reaches 0.4; end-to-end operations research tasks packaged with multi-file operational data, configuration and a submission schema; the agent writes and runs solver code in an isolated Harbor environment (45-minute interaction budget, 5-minute solve limit) and hidden validators score schema and hard-constraint feasibility and normalized objective quality; higher is better. Source: arxiv.org. Saturation forecast: Around April 2027. 14 models tracked.

Top models

#ModelScore
1GPT-5.3 Codex (High)43.9
2Claude Opus 4.639.02
3GPT-5.4 (High)39.02
4Kimi K2.636.59
5GLM-5.131.71
6Claude Sonnet 4.629.27
7GPT-5.4 Mini (High)29.27
8DeepSeek V4 Pro28.46
9DeepSeek V4 Flash26.83
10GLM-524.39
11Qwen 3.6 Plus21.95
12MiMo-V2.5-Pro21.95
13Qwen 3.5 Plus14.63
14MiniMax-M2.72.44

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