OpenAI just updated the GPT-5.5 Instant model, and it's now the default in ChatGPT. The new model brings better real-time accuracy and improved everyday task performance. Here's what changed:
Performance gets a solid boost: hallucination rates are noticeably down in areas like law, finance, and medicine. Image understanding and document parsing have also gotten better.
On style, answers are more concise and to the point — less filler, fewer unnecessary transitions. In short, the new model cuts down on the verbose tendencies that plagued the earlier GPT-5.5. Personalization is improved too.
Rollout started today and is fully live. ChatGPT now uses it by default, though Codex hasn't been updated.
There's also a new feature: memory source visibility. You can now see where your memory comes from through a visual control panel, and if something looks off, you can edit it directly.