MacAgentBench - Productivity: leaderboard

Metric: Pass@1 (%; share of the 224 Productivity tasks solved on a single attempt, deterministic rule-based checks; the benchmark's baseline pure-GUI agent (screenshots and mouse/keyboard actions only), at most 50 steps). Source: arxiv.org. Saturation forecast: Around 2031. 16 models tracked.

Top models

#ModelScore
1GPT-5.457.1
2Qwen 3 VL 235B A22B27.7
3Qwen 3 VL 32B (Thinking)22.3
4Qwen 3 VL 8B (Thinking)21
5Claude Opus 4.618.3
6uitars-1.5-7B18.3
7Gemini 3.1 Pro (Preview)17.9

Interactive version: theaggregate.ai/benchmark?slug=macagentbench-productivity · How It Works · Data refreshed daily, snapshot 2026-09-26.