ERP-Bench: leaderboard
ERP-Bench evaluates agents on long-horizon enterprise ERP workflows, so the input modality is text. The pinned metric is Pass@5 (%), measuring whether the model succeeds within five attempts on the benchmark’s task set. This is a source-attributed result table mirrored by BenchmarkList; the benchmark is produced with the Anchor task-generation pipeline and uses production-grade ERP workflows. Source: BenchmarkList’s attributed result table.
Metric: Pass@5 (%). Source: benchmarklist.com. Saturation forecast: Rough model projection: around 2026. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (Coding) | 73 | |
| 2 | GLM-5.1 (Coding) | 63.3 | |
| 3 | Claude Opus 4.7 (Coding) | 60.7 | |
| 4 | Claude Opus 4.7 (Browser) | 44.3 | |
| 5 | Claude Opus 4.7 (Computer-use) | 38.3 | |
| 6 | GPT-5.5 (Browser) | 23.7 | |
| 7 | Kimi K2.5 (Coding) | 20.3 | |
| 8 | GPT-5.5 (Computer-use) | 16.7 | |
| 9 | GLM-5.1 (Browser) | 12 | |
| 10 | Kimi K2.5 (Browser) | 7.3 |
Interactive version: theaggregate.ai/benchmark?slug=erp-bench · How It Works · Data refreshed daily, snapshot 2026-11-05.