OpenAI is rolling out unlimited free text chats on GPT-5.6 Luna and enhancing safety measures for younger users, while introducing targeted improvements for paid subscribers and addressing reliability concerns.
OpenAI is widening access to ChatGPT while quietly tightening its safety controls for younger users. From this week, free and Go accounts are being moved to GPT-5.6 Luna as the default model, with unlimited text chats for free users due to follow next week. A new Think button will also give harder questions more reasoning time.
The change matters because the limits that remain are specific. OpenAI said the unlimited allowance applies only to text conversations. File uploads, image generation and other tools will keep their existing caps, and abuse safeguards still apply. That means the new offer removes the most visible constraint without turning the service into an entirely unconstrained product.
The move also makes commercial sense. OpenAI cut Luna’s API price sharply on July 30, reducing the cost of input and output tokens to $0.20 and $1.20 per million respectively. Terra also became cheaper, while Sol did not. Axios reported that the pricing changes and wider model rollout are part of OpenAI’s response to growing pressure from lower-cost AI systems, including open-weight models developed in China.
One part of the announcement received less attention but may prove more consequential. OpenAI said it trained the latest release to avoid romantic roleplay, age-restricted challenges and positioning itself as a replacement for real-world relationships for users it believes are under 18. It also said it added age-appropriate limits around sexual content, eating disorders, body-image risks, age-restricted goods, dangerous activity and graphic violence, while encouraging teens to turn to trusted people when they may need support.
For paid users, OpenAI is pushing GPT-5.6 Sol into Chat for Plus and Pro subscribers. The company says the model is designed to give more direct answers, cleaner formatting and less filler, while also correcting users when agreement would be unhelpful. OpenAI has also folded the same model into both instant replies and deeper reasoning, with a new slider on web, mobile and desktop controlling how much effort the system spends on a response.
OpenAI said internal testing found factual errors were much less common in responses from Luna and Sol than in GPT-5.5 Instant, in a narrow evaluation covering financial, medical and legal prompts. Those figures come from the company’s own measurements, so they should be read as a benchmark rather than a full picture of everyday use. Still, they suggest the latest models are being tuned as much for reliability and moderation as for speed.
The update has a practical impact in markets such as India, where paid access can be a meaningful expense. Reuters-style reporting from the announcement suggests the free unlimited text tier will weaken the case for many casual users to subscribe, though Plus still offers Sol, the effort controls and higher limits for files and images. For people who mostly chat, the free tier now looks far more capable. For users who work with documents and media, the paid plan still has a clearer role.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
