GPT-5.5 Instant Health Intelligence Breakthrough: Doctor evaluation surpasses humans, factual question rate drops by 71%

Health is one of the most meaningful application scenarios of ChatGPT, with more than 230 million people getting health information through ChatGPT every week: understanding examination reports, preparing for medical appointments, managing insurance, building healthy habits, etc. OpenAI announces a major breakthrough in the field with GPT-5.5 Instant: HealthBench Professional evaluation reaches the level of cutting-edge Thinking models, doctors rated its answers better than human doctors in 3,500 comparative evaluations, and the rate of factual questions in production traffic dropped by 71% in two months. GPT-5.5 Instant is now available to all ChatGPT free users.

Core competency improvement

GPT-5.5 Instant delivers substantial advances in health: better identifying when urgent care is needed, asking for relevant context, explaining uncertainty, and making complex information easier to understand. In the most demanding health assessments, GPT-5.5 Instant now reaches levels comparable to leading-edge Thinking models.

Key data:

• GPT-5.5 Instant is on par with state-of-the-art models in combined HealthBench and HealthBench Professional evaluations
• Across 3,500 comparative evaluations, doctors rated GPT-5.5 Instant’s answers as superior to human doctors in accuracy, communication, completeness, instruction following, and health decision-making
• Based on privacy-preserving real-time monitoring (billions of messages per week), the factual question rate dropped by 71% in the last two months

Global network of 260 doctors

OpenAI works with 260+ physicians in 60 countries, 49 languages, and 26 specialties. To date, physicians have reviewed more than 700,000 model response samples. On average, a doctor reviews a new answer every few minutes. These review feedback form evaluation criteria to help researchers measure the progress of the model in real health scenarios.

GPT-5.5 Instant is available to all ChatGPT free users (subject to usage restrictions).

GPT-5.5 Instant's breakthroughs in the health field - especially the doctor evaluation of "surpassing human doctors" and the 71% reduction in factual questions - mark that AI has reached a critical node in the "high water mark" field of medical health and is available at the consumer level. The 230 million health consultations per week itself shows that the market demand has been verified.

For the Chinese market, the application of DeepSeek, Doubao, Baidu Wenxin, etc. in health consultation scenarios is still in its early stages. The gap between domestic AI in the medical and health field lies not only in model capabilities, but also in the lack of a "700,000+ doctor review data flywheel" similar to OpenAI - this closed loop of "model capabilities × expert evaluation × continuous iteration" is the core capability that domestic competing products need to focus on building. The global review network of 260 doctors and the HealthBench evaluation system provide a reference model for the quality construction of China's AI medical products.

The follow-up is worthy of attention:

1. Impact of free and open access: Will GPT-5.5 Instant be open to all free users, will it accelerate the popularity of AI health consulting around the world?
2. Domestic AI Medical Benchmarking: How does the evaluation system of Chinese AI companies in the medical and health field benchmark against HealthBench?
3. 260 Doctor Network: Feasibility of an expert review system of similar scale in China
4. Approval and Regulation: Will the improvement of AI health capabilities promote the update of the regulatory framework for medical AI?