Agent Arena | AI Agent Performance Leaderboard

Agent Arena💻Code

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 25, 2026

686,483 sessions

44 models

Pareto Frontier

Net Improvement x Cost/task from real Agent Mode tasks

CtrlScroll

Pareto Optimal Models

Net Improvement x Cost/task

Claude Fable 5.1 (Max)

Anthropic · Proprietary

+15.91%

$10.42/task

GPT 6 Astra (Max)

OpenAI · Proprietary

+14.30%

$6.23/task

Claude Opus 5 (High)

Anthropic · Proprietary

+11.18%

$4.32/task

GPT 6 Sol (Max)

OpenAI · Proprietary

+9.17%

$2.09/task

GPT 5.6 Sol (xHigh)

OpenAI · Proprietary

+6.73%

$1.86/task

Claude Sonnet 5 (High)

Anthropic · Proprietary

+5.68%

$1.39/task

Hy4 preview

Tencent · Apache 2.0

+4.65%

$0.23/task

Deepseek V4.1 Flash (Max)

DeepSeek · MIT

+4.40%

$0.09/task

GPT 5.6 Luna (xHigh)

OpenAI · Proprietary

-0.36%

$0.08/task

Hy3

Tencent · Apache 2.0

-5.49%

$0.05/task

Mimo V2.5 Pro

Xiaomi · MIT

-7.01%

$0.05/task

Cost per Task

Median cost per task (last 14 days)

1Claude Fable 5.1 (Max)$10.42/task+15.91%

2GPT 6 Astra (Max)$6.23/task+14.30%

3Claude Opus 5 (High)$4.32/task+11.18%

4GPT 6 Sol (Max)$2.09/task+9.17%

5Claude Fable 5 (High)$3.88/task+9.11%

6Claude Opus 4.8 (High)$2.65/task+8.27%

7Claude Opus 5 (Max)$7.03/task+8.17%

8GPT 5.6 Sol (xHigh)$1.86/task+6.73%

9Claude Sonnet 5 (High)$1.39/task+5.68%

10GLM 5.2 (Max)$1.89/task+5.51%

11Kimi K3 (Max)$1.54/task+5.23%

12GPT 5.5 (xHigh)$2.12/task+4.69%

13Hy4 preview$0.23/task+4.65%

14Muse Spark 1.3 (Max)$0.74/task+4.60%

15Deepseek V4.1 Flash (Max)$0.09/task+4.40%