OpenAI Updates ChatGPT to GPT‑5.5 Instant, Reducing Hallucinations and Improving Accuracy
Switch ChatGPT’s default model to GPT‑5.5 Instant to cut hallucinations by 52.5 % and improve factuality; test image analysis and web search behavior.
Switch ChatGPT’s default model to GPT‑5.5 Instant and run regression tests on high‑stakes domains to verify reduced hallucinations.
Summary
OpenAI updated ChatGPT’s default model to GPT‑5.5 Instant on May 5 2026, promising smarter, clearer, and more personalized responses. In internal evaluations, GPT‑5.5 Instant produced 52.5 % fewer hallucinated claims than GPT‑5.3 Instant on high‑stakes prompts in medicine, law, and finance, and cut inaccurate claims by 37.3 % on conversations flagged for factual errors. The new model also improves image analysis, STEM question answering, and the decision to use web search for more useful answers. Users report a more natural conversational tone, tighter answers, and better use of shared context for personalization. GPT‑5.5 Instant reduces verbosity, fewer unnecessary follow‑up questions, and eliminates gratuitous emojis, making responses more concise. The update positions GPT‑5.5 Instant as the new default for ChatGPT, replacing GPT‑5.3 Instant across all user interactions.
Key changes
- GPT‑5.5 Instant reduces hallucinated claims by 52.5 % versus GPT‑5.3 Instant on high‑stakes prompts in medicine, law, and finance.
- Inaccurate claims are cut by 37.3 % on conversations flagged for factual errors.
- Image analysis capabilities are improved, allowing better handling of photo and image uploads.
- STEM question answering is enhanced with more accurate responses.
- The model uses web search more judiciously to provide useful answers.
- Conversational tone is more natural and answers are tighter and more personalized.
- Unnecessary follow‑up questions are reduced, and verbosity is trimmed.
- Gratuitous emojis are removed, making responses more concise.