OpenAI GPT-6 Sol & Luna Launch - Reviewer Bypass Attempts: leaderboard
Alignment chart from OpenAI's GPT-6 Sol and GPT-6 Luna launch post (2026-09-22): share of Codex knowledge-work rollouts, starting just after Auto-review denied an action, in which the model retries a substantively similar command or exploits the poorly configured reviewer to bypass it. Percent, lower is better; maximum reasoning effort. The updated version of the Auto-review evaluation, not the Astra launch's circumvention table.
Metric: Bypass attempt rate (%). Source: openai.com. Source: deploymentsafety.openai.com. Source: openai.com. Saturation forecast: Estimated already saturated. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (Max) | 0 | #6 (GPT-6 Astra) |
| 2 | GPT-6.1 Sol (Max) | 0 | #10 (GPT-6.1 Sol) |
| 3 | GPT-6 Sol (Max) | 0 | #24 (GPT-6 Sol) |
| 4 | GPT-6.1 Terra (Max) | 0 | #19 (GPT-6.1 Terra) |
| 5 | GPT-6 Luna (Max) | 0.26 | #93 (GPT-6 Luna) |
| 6 | GPT-5.6 Terra (Max) | 0.6 | #60 (GPT-5.6 Terra) |
| 7 | GPT-5.6 Luna (Max) | 4.26 | #117 (GPT-5.6 Luna) |
| 8 | GPT-5.6 Sol (Max) | 7.31 | #29 (GPT-5.6 Sol) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=openai-gpt-6-sol-luna-launch-reviewer-bypass-attempts · How It Works · Data refreshed daily, snapshot 2026-11-12.