OpenAI has introduced GPT-5.5 Instant as the new default ChatGPT model, claiming significant reductions in hallucinated and inaccurate statements, especially in high-stakes areas like medicine, law, and finance. The upgrade also enhances everyday usability and personalization features.
- Over 50% reduction in hallucinated claims on critical topics
- Improved response precision and contextual awareness
- New memory sources feature for transparency and content control
What happened
OpenAI has rolled out GPT-5.5 Instant as the new default model for ChatGPT. This update follows internal evaluations showing the model produces 52.5% fewer hallucinations than the previous GPT-5.3 Instant, particularly on sensitive subjects such as medicine, law, and finance. It also reduces inaccurate answers by 37.3% on flagged challenging user conversations.
Besides accuracy improvements, GPT-5.5 Instant offers sharper, more concise responses and enhanced capabilities for routine tasks including image analysis and web retrieval when necessary. The rollout began on May 5, 2026, for all ChatGPT users, with GPT-5.3 still available as a fallback option for a limited transition period.
Why it matters
Hallucinations—where AI models fabricate or distort information—have been a persistent challenge for conversational AI, limiting user trust and applicability in professional or critical contexts. A substantial reduction in these errors marks a meaningful advancement in AI reliability, opening doors to broader use cases in legal, financial, and medical domains.
Additionally, GPT-5.5 Instant’s increased accuracy paired with clearer responses helps ensure users receive information that is not only correct but more digestible. Its ability to integrate contextual data from prior chats and linked services like Gmail adds an important layer of personalized interaction that parallels efforts from competitors such as Google's Gemini.
What to watch next
OpenAI plans a phased rollout of enhanced personalization features, starting with Plus and Pro users on the web, soon extending to mobile apps and other subscription tiers including Free, Go, Business, and Enterprise. This expansion will be critical to watch as it determines how effectively the model adapts to diverse user demands on various platforms.
The introduction of the ‘memory sources’ feature is another significant development, providing transparency about which information influenced ChatGPT’s responses and enabling users to delete or correct stored data. How users engage with this feature could influence future data privacy norms and customization standards in AI assistants.