Do Coding Agents Understand Least-Privilege Au: leaderboard
As coding agents gain access to shells, repositories, and user files, least-privilege authorization becomes a prerequisite for safe deployment: an agent should receive enough authority to complete the task, without unnecessary authority that exposes sensitive surfaces. To study whether current models can infer this boundary themselves, we first introduce permission-boundary inference, where a mode
Metric: TSR (self-reported). Source: benchmarklist.com. Saturation forecast: Around July 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 85.8 | #58 |
| 2 | GPT-5 | 76.7 | #98 |
| 3 | Kimi K2.5 | 70 | #144 |
| 4 | GPT-5.3 Codex | 65.8 | #69 |
| 5 | Qwen 3 Coder 480B A35B Instruct | 63.3 | #315 |
| 6 | Claude Opus 4.6 | 61.5 | #63 |
| 7 | GPT-5.4 | 61.1 | #79 |
| 8 | MiniMax-M2.7 | 48.3 | #258 |
| 9 | Qwen 3.5 397B A17B | 45.6 | #146 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=do-coding-agents-understand-least-privilege-au · How It Works · Data refreshed daily, snapshot 2026-10-20.