OpenAI upgrades ChatGPT's default model – GPT-5.5 Instant gets better at understanding context

OpenAI has rolled out significant improvements to GPT-5.5 Instant, the model that powers ChatGPT by default. The company announced that the updated version now excels at understanding conversational context and adapting to user queries as they add more conditions or clarifications. This update builds on the model's initial launch as the default in May, when testing showed it produced 52.5% fewer hallucinated statements and 37.3% fewer factual errors.

The new version focuses on identifying the underlying goal of tasks or questions and carrying context through multiple interactions. GPT-5.5 Instant is now better at processing complex queries, addressing multiple points in a single response. If users clarify their questions or challenge answers, the model adapts more effectively, providing relevant responses instead of repeating previous ones.

Additionally, the upgraded model understands location context more effectively. For example, if asked for restaurant recommendations while traveling, it can surface truly relevant nearby options. GPT-5.5 Instant can also provide product and business information and display images when helpful. OpenAI notes that overall responses will feel more tasteful and cohesive, with natural formatting that moves away from templated chatbot replies.