OpenAI GPT-5 System Card - HealthBench Hard: leaderboard
The "HealthBench Hard" subtask of OpenAI GPT-5 System Card. A benchmark result reported in the OpenAI GPT-5 system card, as self-reported by the model developer.
Metric: Score (%). Source: cdn.openai.com. Saturation forecast: Around January 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 (Thinking) | 46.2 | #98 (GPT-5) |
| 2 | O3 | 31.6 | #124 |
| 3 | GPT-5 | 25.5 | #98 |
| 4 | O4 Mini | 17.5 | #180 |
| 5 | O1 | 7.9 | #183 |
| 6 | GPT-4o | 0 | #343 |
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-system-card-healthbench-hard · How It Works · Data refreshed daily, snapshot 2026-10-20.