OpenAI announced today that it is removing limits on text-based chats for all ChatGPT users.
The company shared the news in a blog post published this morning. It comes shortly after ChatGPT passed 1 billion weekly users.
The update affects every tier of ChatGPT, from free accounts to paid subscriptions. Each group will see different changes rolled out over the coming days.
What Free and Go Users Will Get
Free and Go users will now have GPT-5.6 Luna as their default model. This model replaces GPT-5.5, which previously handled these accounts.
Along with the new model, users will get a "Think" button. This lets people choose more reasoning power when they ask a harder question.
OpenAI said there will still be separate limits for files, images, voice, and image generation. Those limits are not part of this change.
The unlimited text chat feature and the Think button are expected to roll out next week for Free and Go users. Other updates for this group are arriving this week.
Changes Coming for Plus and Pro Users
Plus and Pro subscribers are also getting new tools. They will have access to an upgraded model called GPT-5.6 Sol.
This version is built for quick tasks. That includes things like answering questions, doing web research, giving advice, planning, writing, and helping with decisions.
OpenAI said the new model gives shorter and more reliable answers. The company also clarified that this is a different version from the GPT-5.6 Sol used in Codex and Work products, which remain unchanged.
Plus and Pro users are getting a thinking slider too. This tool lets people adjust how much thought the model puts into a response.
Users can set the slider based on how complex their question is. Simple questions may need less thinking time, while harder ones may need more.
The upgraded GPT-5.6 Sol model is already available to Plus and Pro users as of today. The rest of the updates for this group are expected to follow shortly.
OpenAI shared results from an internal evaluation comparing the new models to GPT-5.5-Instant. Factual errors were 62% lower for GPT-5.6 Luna.
For GPT-5.6 Sol, factual errors dropped by 68% compared to the older model. OpenAI did not release full details on how the evaluation was conducted.
The rollout continues throughout this week and next. Free, Go, Plus, and Pro users are all expected to see the full set of changes in place by the middle of next week.
