GPT-5.5 Instant, the model powering ChatGPT by default for most users, has received a new update from OpenAI with improvements focused on intent recognition and multi-turn conversation coherence. The changes apply automatically across the service without any settings adjustments required.
OpenAI made GPT-5.5 Instant ChatGPT’s default model in May 2026, reporting at the time a 52.5 percent reduction in hallucinated statements during testing and 37.3 percent fewer factual errors compared to its predecessor. The June update builds on that baseline, targeting conversation quality and contextual relevance rather than raw accuracy.
How GPT-5.5 Instant handles context differently now
The central change involves intent recognition. The model now performs better at identifying the underlying goal behind a question rather than responding strictly to its literal phrasing. When a query is ambiguous or contains multiple layers, the model addresses the actual intent more reliably instead of producing a technically accurate but practically unhelpful response.
In multi-turn conversations, context retention has improved substantially. When you add conditions to a question, change direction mid-conversation, or clarify what you meant in a follow-up, the model carries the full conversation history forward. It no longer treats each new message with limited memory of what came before, a pattern that made extended chats increasingly frustrating as they progressed.
For questions containing multiple distinct points, the updated model addresses each part more consistently. Earlier behavior often saw the most prominent point handled well while secondary questions received less attention or were skipped entirely. OpenAI says that has been directly targeted in this update.
Pushback handling has improved as well. If you dispute an answer or refine your question, the model adapts rather than restating the same response with light rewording, which was a recurring source of frustration for regular ChatGPT users.
Location-aware responses and automatic image surfacing
GPT-5.5 Instant now processes location context more precisely. For location-dependent queries such as finding restaurants while traveling, the model takes your actual location into account and returns genuinely nearby results rather than generic suggestions. The same applies to local business information and product recommendations where geographic relevance matters.
OpenAI has also enabled the model to include images in responses without users needing to ask for them explicitly. When it determines that a visual would make an answer more useful, it includes one on its own judgment. The company positions this alongside product recommendations and business information as practical, contextually aware responses rather than purely text-based answers.
Formatting changes
OpenAI says responses from the updated GPT-5.5 Instant should “feel more tasteful and cohesive,” moving away from the heavily templated, bullet-heavy formatting that has defined chatbot output. The aim is output that reads like a considered answer rather than a generated form response.
Because GPT-5.5 Instant is the default model across ChatGPT, none of these changes require any action from users. The updated behavior is live for everyone already using the service.