OpenAI is putting GPT-6 in front of everyone who uses ChatGPT. The rollout began on 7 October for Plus, Pro, Business and Enterprise, with Free and Go a day later. GPT-6 Sol powers the paid tiers, GPT-6 Luna the free ones. OpenAI says ChatGPT reaches more than 1.2 billion people weekly.
Alongside it comes Intelligent UI, which lets the model answer with interactive visuals, buttons, forms, charts and small tools, not text alone. OpenAI also says GPT-6 can begin answering while still thinking: on web-search questions, GPT-6 Instant starts replying 44% sooner on average than GPT-5.6 Instant. Codex and ChatGPT Work are unchanged.
The more useful document is the October safety update published alongside, because it names where the models got worse. On its production benchmarks, GPT-6 Sol regresses significantly on standard self-harm, and Luna on self-harm, gore and sexual content. For under-18 users both regress on age-restricted content, sexual content and emotional reliance, plus an extremism vision test.
Plenty improved. Robustness against prompt injection reaches 99.99% for Sol and 99.79% for Luna, and both beat GPT-5.6 Sol on multiturn jailbreaks at every attacker budget. Both count as High capability in cybersecurity and in biology and chemistry. OpenAI calls the responses behind the regressions generally low in severity, which is its own characterisation, not an outside assessment.
Sources: OpenAI, GPT-6 and Intelligent UI for everyone and GPT-6 Sol and GPT-6 Luna: October 2026 update.