OpenAI to remove ChatGPT text chat caps for Free and Go users
Unlimited text chats are planned for Free and Go accounts
OpenAI says it is removing limits on text-based ChatGPT conversations for all users, after ChatGPT passed 1 billion weekly users. The company plans to make unlimited text chats available next week to its Free and Go tiers.
The experience will use GPT-5.6 Luna as the default model for Free and Go users, replacing GPT-5.5. Those users will also receive a Think button that lets them choose higher reasoning power for complex questions.
The wider access applies specifically to text chats. OpenAI said separate limits will remain in place for files, images, voice and image generation, so the announcement does not turn every ChatGPT capability into an unlimited service.
New model options for paid subscribers
Plus and Pro subscribers are receiving an upgraded GPT-5.6 Sol model today. OpenAI describes it as better suited to quick tasks including questions, web research, advice, planning, writing and decision-making, with more compact and robust answers.
This GPT-5.6 Sol release is separate from the version used for Codex and Work, which OpenAI said is unchanged. Plus and Pro users will also get a thinking slider to adjust how much reasoning the model applies, based on the complexity and number of steps in a query.
In an internal evaluation against GPT-5.5-Instant, OpenAI said factual errors were 62% less common with GPT-5.6 Luna and 68% less common with GPT-5.6 Sol. The figures describe OpenAI’s own comparison and do not remove the need to check important outputs.
What the rollout means for organisations
The rollout extends the direction signalled by GPT-5.6 Luna free access expansion as GPT-5.6 Luna becomes the standard experience for Free and Go users. It also creates a clearer distinction between unrestricted text interaction and capacity-managed features such as uploads, voice and image generation.
For businesses, the practical implication is to set workflow guidance around the remaining feature limits and to retain human review for material decisions, research and factual claims, even where the new models are designed to reduce errors.