AutomationBenchAgents

AutomationBench evaluates whether AI agents can coordinate systems such as CRM, email, and calendars through REST APIs to complete real business workflows. Tasks span sales, marketing, operations, customer service, finance, and HR, requiring agents to discover API endpoints and follow business policies autonomously.

Ranking Sources

#ModelScore

1

DeepSeek-V4.1-FlashDeepSeek

54.8%

2

Gemini 3.7 FlashGoogle

52.3%

3

Qwen3.8-MaxAlibaba · 0902

50.8%

4

Claude Opus 5Anthropic

50.3%

5

Muse Spark 1.3Meta · max mode

49.4%

6

GLM-5.3-FlashZhipu AI

48.8%

7

GLM-5.3Zhipu AI

48.2%

8

GPT-5.6 SolOpenAI · max mode

46.7%

9

Kimi K3Moonshot AI

46.7%

10

Claude Fable 5Anthropic

46.2%

27 models total

Back to benchmarks