OpenAI just rolled out a major update to ChatGPT’s default model, GPT-5.5 Instant, now available to everyone. The headline improvement is a big cut in hallucinations: 52.5% fewer made-up claims in high-stakes areas like medicine, law, and finance compared to GPT-5.3 Instant, with a 37.3% reduction on tricky user-flagged questions. Beyond accuracy, the model delivers tighter, more concise answers without losing personality or substance. It’s better at analyzing photos, handling STEM questions, and deciding when to search the web. The update also supercharges personalization: Instant now uses context from past chats, files, and even Gmail to tailor responses, and it’s faster at finding relevant past conversations. A new feature called memory sources lets you see what context was used to personalize a reply, with controls to delete or correct outdated info. On the Free tier, personalization will draw from a smaller set of past chats. Why it matters: This update shows how small, daily improvements at massive scale — plus real progress on accuracy — can make AI feel less like a tool and more like a genuinely useful assistant you actually want to talk to.