Benchmark
AutomationBench
Published by Zapier over its own automation surface, and cards report different task subsets of it.
- Released
- 2026-03-10
- Categories
- agent
- Openness
- unknown
- Source
- model_reports
- Reported scores
- 5
Reported scores
model_reports
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Claude Opus 5 | Anthropic | 26.0 | 2026-07-24 | evidence |
| DeepSeek-V4-Pro-0813 | DeepSeek | 31.8 | 2026-08-13 | evidence |
| GLM-5.3-Flash | Z.ai | 48.8 | 2026-08-27 | evidence |
| Hy4 preview | Tencent | 32.1 | 2026-08-28 | evidence |
| Kimi K3 | Moonshot AI | 30.8 | 2026-06-13 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.