OpenAI began rolling out GPT-6 for ChatGPT on Wednesday, with Plus, Pro, Business and Enterprise subscribers receiving a version called GPT-6 Sol. Free and Go users will gain access to the new model, GPT-6 Luna, starting October 8, according to OpenAI's announcement and a post on the company's X account.

The rollout is worldwide and takes place in the Chat tab of ChatGPT, with workplace admin settings controlling enterprise access — so the timing for workplace teams rests with their administrators. OpenAI said both models are “tuned for everyday conversation.” This release makes no changes to the models powering ChatGPT's Work and Codex experiences.

Sol reaches paid tiers first

OpenAI pitched GPT-6 to ChatGPT's more than 1.2 billion weekly users, with the staggered schedule deciding when each segment of that audience — paying subscribers first, then free and Go users from October 8 — gets the new models. The first GPT-6 models reached paid customers last month, and Sol and Luna have already been available to developers. OpenAI cut API prices for both models by at least 50% compared with their GPT-5.6 predecessors on September 22, as Cryptopolitan reported.

Sol is priced at $2 per million input tokens and $10 per million output tokens, down from $4 and $20 for GPT-5.6 Sol. Luna costs $0.10 and $0.50, versus $0.20 and $1.20 for GPT-5.6 Luna.

OpenAI's Dots agents, launched September 29, run on a third model, GPT-6 Astra. The company said GPT-6 builds on several of Astra's safety advances.

Intelligent UI arrives as search answers start 44% sooner

GPT-6 also introduces Intelligent UI. ChatGPT answers can now blend graphics, tappable buttons, forms, charts and interactive elements, with the format determined by the question. A comparison can appear side by side, an explanation can take the form of an interactive diagram, and simple questions still receive plain text.

Users can also ask ChatGPT to build a tool during the conversation. OpenAI's examples include a savings calculator, a bill splitter for dinner with friends and a game. The company built a library of native streamable components and a compiler that processes the interface as GPT-6 writes it, allowing the interface to appear gradually rather than waiting for the full response to complete.

The interactive-answer capability brings ChatGPT in line with Gemini, where Google launched the comparable Neural Expressive feature in May.

GPT-6 can additionally start answering while it is still thinking, building on the answer as it goes. For web search questions, OpenAI said GPT-6 Instant begins responding 44% sooner on average than GPT-5.6 Instant. In internal tests on difficult problems, GPT-6 more often addressed the key aspect of a user's question than GPT-5.6. The company also said GPT-6 Extra High starts answering as fast as GPT-5.6 Medium on everyday agentic tasks while scoring better than GPT-5.6 Extra High.

Expanded safety training

OpenAI also increased safety training for GPT-6 against misuse involving cyberattacks, biological threats and violence, a step that comes as the models prepare to reach ChatGPT's free and Go tiers on October 8. In adversarial testing, the company said GPT-6 was more resistant to attempts to circumvent that training, particularly attacks that adapt over multiple turns.