APEX-AgentsAgents

APEX-Agents evaluates AI agents on long-horizon, cross-application tasks. It contains 480 realistic workplace scenarios designed by investment bankers, management consultants, and corporate lawyers. Agents must navigate file and tool environments autonomously and produce multi-step deliverables. It uses Pass@1 scoring.

Ranking Sources

#ModelScore

1

Claude Fable 5Anthropic

59.2%

2

Grok 4.6xAI

57.5%

3

GPT-5.6 SolOpenAI

56.7%

4

Grok 4.5xAI

47.1%

5

Claude Opus 4.8Anthropic

42.5%

6

GPT-5.5OpenAI

41.7%

7

Kimi K3Moonshot AI

41.0%

8

Claude Opus 4.7Anthropic

37.2%

9

GLM-5.2Zhipu AI

35.6%

10

Seed2.1 ProByteDance

33.8%

24 models total

Back to benchmarks