OpenAI has officially launched GPT-5.5 Instant, which will serve as the default model for ChatGPT, succeeding GPT-5.3 Instant. This new model aims to enhance accuracy and minimize hallucinations in critical sectors such as law, medicine, and finance, while maintaining speed. Following the general release of the GPT-5.5 series last month, which introduced significant advancements in code generation and reasoning, GPT-5.5 Instant has shown substantial performance improvements. Notably, it achieved 81.2 points on the AIME 2025 math benchmark, a marked increase from the previous score of 65.4. Additionally, it scored 76 on the MMMU-Pro multimodal reasoning benchmark, up from 69.2. One of the key updates is the introduction of context-related improvements, enabling the AI to utilize integrated search tools to reference prior conversations and user-uploaded files, enhancing the personalization of responses. Initially, this feature will be available to Plus and Pro subscribers through the website, with a mobile app update expected soon. Furthermore, GPT-5.5 Instant will be accessible to developers via API under the name 'chat-latest,' while GPT-5.3 will remain available for paid users for a limited transition period of three months.