Agent Arena | AI Agent Performance Leaderboard

Agent Arena🏆Overall

Dynamic ranking of models on how well they orchestrate tools for real-world agentic tasks, based on signals like tool reliability, task completion, and steerability.

Sep 1, 2026

2,188,416 sessions

58 models

Pareto Frontier

Net Improvement x Cost/task from real Agent Mode tasks

CtrlScroll

Pareto Optimal Models

Net Improvement x Cost/task

Claude Opus 5 (High)

Anthropic · Proprietary

+13.74%

$2.43/task

Claude Fable 5 (High)

Anthropic · Proprietary

+10.61%

$2.37/task

GPT 5.6 Sol (xHigh)

OpenAI · Proprietary

+9.49%

$1.23/task

Kimi K3 (Max)

Moonshot · Kimi K3 license

+8.71%

$0.80/task

GLM 5.2 (Max)

Z.ai · MIT

+6.23%

$0.47/task

Grok 4.5

SpaceXAI · Proprietary

+6.17%

$0.41/task

Hy4 preview

Tencent · Apache 2.0

+5.97%

$0.26/task

Gemini 3.8 Flash (High)

Google · Proprietary

+5.94%

$0.22/task

GLM 5.3 Flash

Z.ai · MIT

+4.03%

$0.12/task

GPT 5.6 Luna (xHigh)

OpenAI · Proprietary

+2.15%

$0.08/task

Hy3

Tencent · Apache 2.0

-1.51%

$0.06/task

Mimo V2.5 Pro

Xiaomi · MIT

-3.57%

$0.05/task

Cost per Task

Median cost per task (last 14 days)

1Claude Opus 5 (High)$2.43/task+13.74%

2Claude Opus 5 (Max)$3.82/task+11.69%

3Claude Fable 5 (High)$2.37/task+10.61%

4GPT 5.6 Sol (xHigh)$1.23/task+9.49%

5Claude Opus 4.8 (High)$1.65/task+9.22%

6Kimi K3 (Max)$0.80/task+8.71%

7GPT 5.5 (xHigh)$1.05/task+7.53%

8Claude Sonnet 5 (High)$0.87/task+7.51%

9Claude Opus 4.7 (High)$1.09/task+6.49%

10GLM 5.2 (Max)$0.47/task+6.23%

11Claude Opus 4.7$1.04/task+6.18%

12Grok 4.5$0.41/task+6.17%

13Hy4 preview$0.26/task+5.97%

14GPT 5.5 (High)$0.82/task+5.97%

15Gemini 3.8 Flash (High)$0.22/task+5.94%