Workspace-Bench — leaderboard
Workspace-agent benchmark over file-heavy tasks involving documents, spreadsheets, presentations, code, and multi-file dependencies.
Metric: Overall Score (%). Source: workspace-bench.github.io. 45 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | OpenClaw (Opus-4.7, Test-Rubrics-Checked) | 66.7 |
| 2 | ClaudeCode (Opus-4.7, Test-Rubrics-Checked) | 64.7 |
| 3 | Hermes (Opus-4.7, Test-Rubrics-Checked) | 64.5 |
| 4 | DeepAgent (GLM-5.1, Test-Rubrics-Checked) | 61 |
| 5 | Hermes (GLM-5.1, Test-Rubrics-Checked) | 57.7 |
| 6 | OpenClaw (GLM-5.1, Test-Rubrics-Checked) | 57.5 |
| 7 | OpenClaw (Qwen-3.6-Plus, Test-Rubrics-Checked) | 55.6 |
| 8 | ClaudeCode (MiniMax-M2.7, Test-Rubrics-Checked) | 54.6 |
| 9 | DeepAgent (Opus-4.7, Test-Rubrics-Checked) | 54.4 |
| 10 | Codex (GLM-5.1, Test-Rubrics-Checked) | 53.1 |
Interactive version: theaggregate.ai/benchmark?slug=workspace-bench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.