OpenAI GPT-6 Sol & Luna Launch - Coding Deception: leaderboard
Alignment chart from OpenAI's GPT-6 Sol and GPT-6 Luna launch post (2026-09-22): share of final responses on coding tasks with any misleading claim, such as false reports of completed actions, tool access, verification or background work. The tasks are chosen to elicit dishonesty. Percent, lower is better; maximum reasoning effort. The updated version the GPT-6 Astra card's 2026-09-22 change log describes.
Metric: Deception rate (%). Source: openai.com. Saturation forecast: Estimated already saturated. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (Max) | 0.51 | #5 (GPT-6 Astra) |
| 2 | GPT-6 Sol (Max) | 1.3 | #17 (GPT-6 Sol) |
| 3 | GPT-6 Luna (Max) | 2.81 | #62 (GPT-6 Luna) |
| 4 | GPT-5.6 Luna (Max) | 9.54 | #80 (GPT-5.6 Luna) |
| 5 | GPT-5.6 Sol (Max) | 10.41 | #22 (GPT-5.6 Sol) |
No result here: #3 Claude Opus 5.5, #6 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=openai-gpt-6-sol-luna-launch-coding-deception · How It Works · Data refreshed daily, snapshot 2026-11-07.