ApprenticeBench - GUI (Computer Use): leaderboard
NeoCognition's ApprenticeBench: an agent joins a simulated construction company as an accounts-payable apprentice and processes 100 vendor bills in order in an Odoo ERP with only the handbook, six months of history and mentor feedback, which is frequent in the first month and sparse afterwards. This track uses the GUI through computer-use tools inside a Codex or Claude Code harness; score is the cumulative share of bills passed, higher is better.
Metric: Cumulative Success Rate (%). Source: neocognition.io. Saturation forecast: Around June 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5.1 (Max) | 72 | #10 (Claude Fable 5.1) |
| 2 | GPT-6 Astra (Max) | 68 | #6 (GPT-6 Astra) |
| 3 | GPT-6 Astra (xHigh) | 61 | #6 (GPT-6 Astra) |
| 4 | Claude Fable 5.1 (xHigh) | 59 | #10 (Claude Fable 5.1) |
| 5 | Claude Fable 5 (xHigh) | 43 | #20 (Claude Fable 5) |
| 6 | Claude Opus 5 (Max) | 36 | #16 (Claude Opus 5) |
| 7 | Claude Fable 5 (Max) | 34 | #20 (Claude Fable 5) |
| 8 | GPT-5.6 Sol (Max) | 26 | #22 (GPT-5.6 Sol) |
| 9 | Gemini 3.8 Flash | 24 | #30 |
| 10 | Muse Spark 1.3 (Max) | 19 | #31 (Muse Spark 1.3) |
| 11 | Kimi K3 (Max) | 18 | #35 (Kimi K3) |
| 12 | Gemini 3.7 Flash | 16 | #24 |
| 13 | GPT-5.6 Terra (Max) | 16 | #40 (GPT-5.6 Terra) |
| 14 | Claude Sonnet 5 (Max) | 16 | #65 (Claude Sonnet 5) |
| 15 | GPT-5.6 Luna (Max) | 7 | #79 (GPT-5.6 Luna) |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=apprenticebench-gui-computer-use · How It Works · Data refreshed daily, snapshot 2026-11-05.