A New Era of Accuracy and Reliability for ChatGPT
OpenAI has officially launched GPT-5.5 Instant, establishing it as the new default model for its popular conversational AI, ChatGPT. This significant upgrade aims to deliver a more dependable and accurate experience for everyday users, succeeding the previous GPT-5.3 Instant version. The rollout, which began this week, makes the model globally available and accessible to developers via the API under "chat-latest".
A primary focus of GPT-5.5 Instant is a substantial reduction in "hallucinations," or factually incorrect claims, especially in critical fields. OpenAI's internal evaluations show that the new model produced 52.5% fewer hallucinated claims than its predecessor on high-stakes prompts in areas such as medicine, law, and finance. Furthermore, it reduced inaccurate claims by 37.3% in challenging conversations previously flagged by users for factual errors. The model also demonstrated improved logical reasoning, scoring 81.2 on the AIME 2025 math test, a notable increase from GPT-5.3 Instant's 65.4.