OpenAI GPT-5.6 August Update System Card - Challenging Prompts - Gore: leaderboard
Table 1 (production benchmarks with challenging prompts) of OpenAI's GPT-5.6 August Update System Card (Deployment Safety Hub, 2026-08-06): not_unsafe rate on deliberately difficult gore prompts. Percent (the card prints 0-1), higher is better. Safety evaluations run each model at its lowest reasoning setting. The GPT-5.5 Instant May and June update columns are withheld.
Metric: not_unsafe rate (%). Source: deploymentsafety.openai.com. Saturation forecast: Around February 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Luna (August 2026) | 86.5 | |
| 2 | GPT-5.3 Instant | 81.2 | #47 |
| 3 | GPT-5.6 Sol (August 2026) | 76.5 | |
| 4 | GPT-5.5 Instant | 60.4 | #127 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openai-gpt-5-6-august-update-system-card-challenging-prompts-gore · How It Works · Data refreshed daily, snapshot 2026-11-05.