SAN FRANCISCO, CALIFORNIA / RankWire.AI / – As AI services continue to evolve, OpenAI is expanding access to ChatGPT by removing text chat restrictions for Free and Go users under the GPT-5.6 Luna update. The company announced that unlimited text interactions will start next week, with safeguards in place to prevent misuse. During this rollout, Luna has become the default model for these plans. Additionally, Free users will soon see a new Think button, enabling the model to dedicate extra time for reasoning on more complex queries before providing a response.
The new policy applies specifically to standard text-based conversations and does not extend to features like image generation, file uploads, or other tools that require additional computational resources. OpenAI confirmed that the Think button will be introduced alongside unlimited text access next week, allowing users to activate it when a question demands more detailed thought. This update allows Free and Go accounts to enjoy more of ChatGPT’s core conversational capabilities, while separate limits remain in place for other functionalities.
GPT-5.6 includes several models, such as Luna, Sol, and Terra, each tailored for different performance needs. Luna is designed as the faster and more affordable option within this lineup. According to OpenAI, approximately one billion people use ChatGPT every week for activities like research, planning, writing, and searching. With this recent update, Luna now serves as the primary model for Free and Go users, while other GPT-5.6 variants remain available across various products and plans optimized for different tasks and access levels.
Free access is now broader, but tool restrictions remain
Subscribers to ChatGPT Plus and Pro are receiving an upgraded GPT-5.6 Sol model that supports both everyday use and more intricate reasoning. The company notes that Sol provides more targeted responses and reduces unnecessary details when shorter replies suffice. Paid tiers also gain a reasoning slider accessible on web, mobile, and desktop platforms, allowing users to customize how much processing power the model applies to each request. This feature is useful for tasks like research, coding, writing, and planning, which may require varying levels of reasoning.
OpenAI shared results from internal testing focused on the factual accuracy of its newest models. The evaluations examined responses to financial, medical, and legal questions that demanded detailed factual accuracy. The company reported a roughly 62% decrease in answers with at least one factual error when comparing GPT-5.6 Luna to GPT-5.5 Instant. Meanwhile, GPT-5.6 Sol showed an improvement of about 68% under similar testing conditions. OpenAI also highlighted enhancements in handling dates, numbers, sources, rules, and assumptions, especially when these details influenced the accuracy of responses.
New models incorporate additional protections for younger users
The GPT-5.6 models now feature enhanced safeguards designed to protect users believed to be under 18. These measures restrict responses involving romantic roleplay, age-restricted challenges, and claims that might encourage users to replace real-world relationships. Additional limits are in place for content related to sexual material, eating disorders, dangerous activities, restricted products, and graphic violence. OpenAI explained that it has implemented further assessments for responses involving younger users. The same abuse prevention measures will continue to be enforced when unlimited text conversations are available to Free and Go accounts through the expanded ChatGPT service.
This rollout impacts the core ChatGPT service but does not alter every product utilizing GPT-5.6. Versions of GPT-5.6 Sol used in ChatGPT Work and Codex remain unaffected by this update. Users with Plus and Pro subscriptions can now access the updated Sol model along with the new reasoning controls. Meanwhile, Free and Go users will see Luna as their default model this week. The introduction of unlimited text conversations and the Think button is scheduled for next week, with separate restrictions remaining on image generation, uploads, and other resource-heavy features.