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 8, 2026

1,543,817 sessions

43 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

+14.51%

$4.46/task

GPT 6 Astra (Max)

OpenAI · Proprietary

+12.55%

$4.06/task

Claude Opus 5 (High)

Anthropic · Proprietary

+11.42%

$2.45/task

Claude Fable 5 (High)

Anthropic · Proprietary

+9.15%

$2.36/task

GPT 5.6 Sol (xHigh)

OpenAI · Proprietary

+7.75%

$1.21/task

Kimi K3 (Max)

Moonshot · Kimi K3 license

+6.59%

$0.81/task

Hy4 preview

Tencent · Apache 2.0

+5.37%

$0.25/task

DeepSeek V4 Pro (High) (0813)

DeepSeek · MIT

+4.43%

$0.23/task

Gemini 3.8 Flash (High)

Google · Proprietary

+4.01%

$0.17/task

Deepseek V4 Flash (High) (20260731)

DeepSeek · MIT

+2.22%

$0.16/task

GLM 5.3 Flash

Z.ai · MIT

+2.00%

$0.11/task

GPT 5.6 Luna (xHigh)

OpenAI · Proprietary

+1.11%

$0.08/task

Hy3

Tencent · Apache 2.0

-4.22%

$0.05/task

Mimo V2.5 Pro

Xiaomi · MIT

-5.08%

$0.05/task

Cost per Task

Median cost per task (last 14 days)

1Claude Fable 5.1 (Max)$4.46/task+14.51%

2GPT 6 Astra (Max)$4.06/task+12.55%

3Claude Opus 5 (High)$2.45/task+11.42%

4Claude Opus 5 (Max)$3.54/task+10.79%

5Claude Fable 5 (High)$2.36/task+9.15%

6GPT 5.6 Sol (xHigh)$1.21/task+7.75%

7Claude Opus 4.8 (High)$1.44/task+7.69%

8Kimi K3 (Max)$0.81/task+6.59%

9Claude Sonnet 5 (High)$0.88/task+6.31%

10Hy4 preview$0.25/task+5.37%

11GPT 5.5 (xHigh)$1.06/task+5.33%

12GLM 5.2 (Max)$0.64/task+4.64%

13DeepSeek V4 Pro (High) (0813)$0.23/task+4.43%

14Gemini 3.8 Flash (High)$0.17/task+4.01%

15Grok 4.5$0.50/task+3.93%