OpenAI GPT-5.5 Launch - Humanity's Last Exam (with tools): leaderboard
The "Humanity's Last Exam (with tools)" subtask of OpenAI GPT-5.5 Launch. A benchmark result reported in the OpenAI GPT-5.5 launch announcement, as self-reported by the model developer.
Metric: Score (%). Source: openai.com. Source: www-cdn.anthropic.com. Source: www-cdn.anthropic.com. Source: openai.com. Saturation forecast: Around March 2028. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Mythos Preview | 64.7 | #4 |
| 2 | Claude Mythos 5 | 64.5 | #2 |
| 3 | GPT-5.4 Pro (xHigh) | 58.7 | #32 (GPT-5.4 Pro) |
| 4 | Claude Opus 4.8 | 57.9 | #37 |
| 5 | GPT-5.5 Pro | 57.2 | #17 |
| 6 | Claude Opus 4.7 | 54.7 | #49 |
| 7 | GPT-5.5 | 52.2 | #31 |
| 8 | GPT-5.4 | 52.1 | #79 |
| 9 | Gemini 3.1 Pro (Preview) | 51.4 | #58 |
| 10 | GPT-5.2 Pro | 50 | #61 |
| 11 | GPT-5.2 | 45.5 | #110 |
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-5-5-launch-humanity-s-last-exam-with-tools · How It Works · Data refreshed daily, snapshot 2026-10-20.