Google Carbon Report - ExploitBench - Autonudge capabilities: leaderboard
Google DeepMind Carbon safety report, Figure 7.1. 41 tasks x 16 capabilities=656; ACE grants full16. Autonudge,300-turn limit,best-of-three union of achieved capabilities; internal non-cyber-specialized Antigravity + unmodified official MCP servers. Metric: Capabilities reached (%). Higher is better. Source: https://storage.googleapis.com/deepmind-media/gemini/gemini_4_carbon_safety_report.pdf#page=89
Metric: Capabilities reached (%). Source: storage.googleapis.com. Saturation forecast: Around February 2027. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Carbon | 97 | #2 |
| 2 | Gemini 4 Argon | 87 | #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-exploitbench-autonudge-capabilities · How It Works · Data refreshed daily, snapshot 2026-11-12.