OpenAI GPT-6 Sol & Luna Launch - Factual Errors on User-Flagged Conversations: leaderboard
Chart from OpenAI's GPT-6 Sol and GPT-6 Luna launch post (2026-09-22): share of answers with any factual error on OpenAI's internal factuality evaluation, built from de-identified ChatGPT conversations in which users flagged a factual error by an earlier model. Percent, lower is better. Each model at five reasoning efforts, Low to Max; not length-controlled. The prompts are chosen to induce errors, so rates are far above typical use.
Metric: Answers with any factual error (%). Source: openai.com. Source: openai.com. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (High) | 3.9 | #5 (GPT-6 Astra) |
| 2 | GPT-6 Astra (Max) | 3.91 | #5 (GPT-6 Astra) |
| 3 | GPT-6 Astra (xHigh) | 3.99 | #5 (GPT-6 Astra) |
| 4 | GPT-6.1 Sol (xHigh) | 4.12 | #9 (GPT-6.1 Sol) |
| 5 | GPT-6 Astra (Medium) | 4.41 | #5 (GPT-6 Astra) |
| 6 | GPT-6.1 Sol (High) | 4.52 | #9 (GPT-6.1 Sol) |
| 7 | GPT-6 Sol (xHigh) | 4.52 | #22 (GPT-6 Sol) |
| 8 | GPT-6 Sol (Max) | 4.57 | #22 (GPT-6 Sol) |
| 9 | GPT-6.1 Sol (Max) | 4.61 | #9 (GPT-6.1 Sol) |
| 10 | GPT-6 Sol (High) | 5.14 | #22 (GPT-6 Sol) |
| 11 | GPT-6 Astra (Low) | 6.26 | #5 (GPT-6 Astra) |
| 12 | GPT-6.1 Sol (Medium) | 6.29 | #9 (GPT-6.1 Sol) |
| 13 | GPT-6 Sol (Medium) | 6.86 | #22 (GPT-6 Sol) |
| 14 | GPT-6 Luna (Max) | 7.56 | #90 (GPT-6 Luna) |
| 15 | GPT-6.1 Sol (Low) | 7.72 | #9 (GPT-6.1 Sol) |
No result here: #4 Claude Opus 5.5, #10 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=openai-gpt-6-sol-luna-launch-factual-errors-on-user-flagged-conversations · How It Works · Data refreshed daily, snapshot 2026-11-08.