OpenAI brings GPT-6 and Intelligent UI to all ChatGPT users

OpenAI's rollout caps a gradual build-up. The company put GPT-6 family models in front of paying users in September, but this release is the first aimed at the whole ChatGPT audience, spanning Plus, Pro, Business, and Enterprise plans on October 7 and free and Go plans the next day. The new models, GPT-6 Sol and GPT-6 Luna, slot in as replacements for GPT-5.6 Sol and GPT-5.6 Luna, and carry over safety improvements from GPT-6 Astra.

The most visible change is Intelligent UI, which lets GPT-6 decide whether an answer should be text, a visual, or something the user can actually operate. In an official example comparing two models on a roast-lamb dinner plan, GPT-5.6 Instant returned a text portion table and schedule, while GPT-6 Instant showed menu images, a feature that recalculates ingredient amounts when the guest count changes, and a cooking checklist. OpenAI also says GPT-6 can start writing an answer while it is still thinking, rather than making users wait for the reasoning to finish.

Safety is the quieter story. Under OpenAI's Preparedness Framework, both models are treated as having "High" capability in cybersecurity and biochemical fields, with the same safeguards applied to GPT-5.6, though they do not reach the "High" bar on AI self-improvement. Some safety evaluations did dip — statistically significant drops for self-harm for both Sol and Luna, and for violent and sexual content on Luna — which OpenAI says it reviewed by hand and judged borderline and low-severity. The commercial test now is whether interactive answers feel genuinely useful rather than gimmicky, and whether enterprise administrators switch the feature on.