AutomationBench (Operations)
Agents · 2026-04-20
This AutomationBench subtrack evaluates agents on end-to-end operations workflows built with Zapier applications. Scores report the percentage of tasks completed successfully.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 56.0 |
| Gemini 3.6 Flash | 55.0 |