AARRI-Bench (Mini-SWE-Agent) - Context: leaderboard

Metric: Task success rate (%) under the 0/1 reward on the context tasks: sensitivity to the state of a field (core contributions, data validity, genuine advances); Harbor containerized runs, Mini-SWE-Agent agent harness; a configuration's tasks without a valid run are left out of its denominator; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 8 models tracked.

Top models

#ModelScore
1Claude Opus 4.764.7
2Kimi K2.659.4
3MiniMax-M2.755.9
4DeepSeek V4 Flash50
5Qwen 3.6 Plus44.1
6Qwen 3 235B A22B 2507 (Thinking)41.2
7GPT-OSS-120B33.3
8Qwen 3 Next 80B A3B Instruct29.4

Interactive version: theaggregate.ai/benchmark?slug=aarri-bench-mini-swe-agent-context · How It Works · Data refreshed daily, snapshot 2026-09-29.