Google Carbon Report - Privacy Safety Security Advice - Mean committed pitfalls: leaderboard
Google DeepMind Carbon safety report, Figure 5.17. Same 450-question privacy/safety/security advice suite. Count of committed high-severity pitfalls per response; 32 predefined pitfalls. Not overall helpfulness. Metric: Mean committed pitfalls per response. Lower is better. Source: https://storage.googleapis.com/deepmind-media/gemini/gemini_4_carbon_safety_report.pdf#page=80
Metric: Mean committed pitfalls per response. Source: storage.googleapis.com. Saturation forecast: Around 2030. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Carbon | 1.96 | #2 |
| 2 | Gemini 4 Argon | 1.32 | #11 |
| 3 | Gemini 3.8 Flash | 1.22 | #36 |
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-privacy-safety-security-advice-mean-committed-pitfalls · How It Works · Data refreshed daily, snapshot 2026-11-12.