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 28, 2026

2,046,422 sessions

46 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

+14.06%

$3.54/task

Claude Opus 5.5 (High)

Anthropic · Proprietary

+11.84%

$1.41/task

GPT 6 Sol (Max)

OpenAI · Proprietary

+8.80%

$0.81/task

Claude Sonnet 5 (High)

Anthropic · Proprietary

+4.80%

$0.74/task

Hy4 preview

Tencent · Apache 2.0

+4.10%

$0.25/task

Deepseek V4.1 Flash (Max)

DeepSeek · MIT

+3.86%

$0.08/task

GPT 6 Luna (Max)

OpenAI · Proprietary

+1.59%

$0.05/task

Hy3

Tencent · Apache 2.0

-5.75%

$0.04/task

Mimo V2.5 Pro

Xiaomi · MIT

-6.61%

$0.03/task

Cost per Task

Median cost per task (last 14 days)

1Claude Fable 5.1 (Max)$3.54/task+14.06%

2Claude Opus 5.5 (High)$1.41/task+11.84%

3GPT 6 Astra (Max)$2.75/task+10.36%

4Claude Opus 5 (Max)$3.09/task+9.54%

5Claude Opus 5 (High)$2.27/task+9.38%

6GPT 6 Sol (Max)$0.81/task+8.80%

7Claude Fable 5 (High)$1.71/task+7.99%

8Claude Opus 4.8 (High)$1.14/task+6.87%

9GPT 5.6 Sol (xHigh)$0.88/task+6.26%

10Claude Sonnet 5 (High)$0.74/task+4.80%

11GPT 5.5 (xHigh)$1.11/task+4.24%

12Hy4 preview$0.25/task+4.10%

13Kimi K3 (Max)$0.66/task+4.03%

14Grok 4.7 (xHigh)$1.24/task+4.02%

15Deepseek V4.1 Flash (Max)$0.08/task+3.86%