OmniaBench - DAG: leaderboard

Metric: Pass@1 (%; 354 multi-turn stateful tasks built from sampled tool-dependency graphs, one per level-2 domain, with a simulated user; a task counts as solved only when every rubric item is satisfied (Program tasks by a VerifyCode check); up to 200 tool-call steps; challenging set of 644 tasks evaluated June 29 to July 15, 2026). Source: arxiv.org. Saturation forecast: Around June 2028. 22 models tracked.

Top models

#ModelScore
1Claude Sonnet 5 (Max)57.34
2GPT-5.6 Sol (High)55.37
3GLM-5.2 (Max)54.8
4GPT-5.5 (High)54.8
5Gemini 3.5 Flash (High)54.24
6Claude Opus 4.7 (Max)53.39
7DeepSeek V4 Pro (Max)52.54
8DeepSeek V4 Flash (Max)51.13
9GLM-5 (Thinking)50.85
10Qwen 3.5 397B A17B50.56
11Gemini 3.1 Pro (Preview)50.28
12Kimi K2.6 (Thinking)49.72
13Qwen 3.7 Max (Thinking)48.59
14Qwen 3.6 27B48.02
15Qwen 3.6 35B A3B43.22

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