GPT-5.5 Instant Update: Improved Style, Readability, & Accuracy

OpenAI has updated GPT-5.5 Instant, improving its response style, readability, and personalization. This update, released on May 28, boosts the model's performance on various benchmarks. For example, it scored 81.2 on the AIME 2025 math test, up from 65.4. It also achieved 76 on the MMMU-Pro benchmark, compared to 69.2 for the prior model. Internal testing showed a 52.5% reduction in hallucinated claims and 37.3% fewer inaccurate claims in flagged conversations. The update also expands an API route and introduces new personalization features, like searchable past conversations and a "memory sources" user interface. These features are rolling out to paid users first, with wider availability coming soon. The bottom line is that these changes aim to make AI interactions more accurate and personalized for users.