Google Carbon Report - ExploitGym - 6h: leaderboard
Google DeepMind Carbon safety report, Figure 7.2. 869 challenges; one attempt/task; Antigravity; no internet; benchmark mitigations off; 6-hour timeout. Success requires target flag and LLM confirmation of intended vulnerability. Metric: Intended exploit rate (%). Higher is better. Source: https://storage.googleapis.com/deepmind-media/gemini/gemini_4_carbon_safety_report.pdf#page=90
Metric: Intended exploit rate (%). Source: storage.googleapis.com. Saturation forecast: Around March 2028. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Carbon | 53.5 | #2 |
| 2 | Gemini 4 Argon | 39.5 | #11 |
No result here: #1 Claude Fable 5.2, #5 Claude Opus 5.5, #6 GPT-6 Astra.
Interactive version: theaggregate.ai/benchmark?slug=google-carbon-report-exploitgym-6h · How It Works · Data refreshed daily, snapshot 2026-11-12.