Agent Arena - Confirmed Success: leaderboard

Metric: Confirmed Success (%). Source: arena.ai. 58 models tracked.

Top models

#ModelScore
1Kimi K3 (Max)16.9
2Claude Opus 5 (Max)15.49
3Hy4 preview15.04
4Claude Opus 5 (High)14.96
5GLM-5.3 Flash14.56
6GLM-5.3 (Max)12.36
7Qwen3.8-Flash-Next11.86
8Gemini 3.8 Flash (High)11.12
9Grok 4.6 (xHigh)10.6
10Gemini 3.7 Flash (High)9.8
11GLM-5.2 (Max)8.65
12Qwen 3.8 Max8.38
13GPT-5.6 Sol (xHigh)7.57
14Qwen 3.8 27B7.08
15Claude Fable 5 (High)6.81

Interactive version: theaggregate.ai/benchmark?slug=agent-arena-confirmed-success · How It Works · Data refreshed daily, snapshot 2026-09-05.