OpenAI、GPT-5.5 InstantをChatGPTのデフォルトモデルに
OpenAIはGPT-5.5 Instantをリリースし、GPT-5.3 Instantに代わってChatGPTのデフォルトモデルとした。新モデルはAIME 2025数学で65.4から81.2にスコアを向上させ、法律、医療、金融分野でのハルシネーション削減を重視している。出典引用や編集を可能にするコンテキスト管理機能が、PlusおよびProユーザーに現在展開中だ。
OpenAI released GPT-5.5 Instant, replacing GPT-5.3 Instant as the default ChatGPT model. The new model improves performance on the AIME 2025 math benchmark to 81.2 (up from 65.4) and the MMMU-Pro multimodal benchmark to 76.0 (up from 69.2). The release focuses on reducing hallucinations in sensitive sectors like law and medicine while maintaining low latency. The model introduces context management features, allowing it to reference past conversations, Gmail, and files to provide personalized answers. This citation feature is available immediately to Plus and Pro web users, with mobile and enterprise access planned for the coming weeks. Developers can access the model via the API as 'chat-latest', while the API for GPT-5.3 will remain available for paid users for three months.