OpenAI Removes Text Chat Limits for ChatGPT Free and Go Users, Upgrades GPT-5.6 Sol for Plus and Pro

OpenAI is removing text chat rate limits for ChatGPT Free and Go tier users starting next week, according to the company. Currently, users on these tiers can hit rate limits after a certain number of text-only exchanges in a given period. That cap disappears for text conversations specifically — messages involving file uploads or images will still be subject to limits.

Alongside the removal of text limits, OpenAI is adding a "Think" button for Free and Go users next week, letting them opt into higher reasoning for harder questions on demand rather than relying solely on the default model's automatic routing. This week, OpenAI is also upgrading the default model for Free and Go users to GPT-5.6 Luna, the company's most recently released model in that tier.

Changes for Plus and Pro subscribers

Paying subscribers are getting a different update. OpenAI says GPT-5.6 Sol — the model currently powering ChatGPT for Plus and Pro users — is being revised to be "more reliable with facts." According to OpenAI, the updated model is "designed to make fewer mistakes — especially when answers depend on dates, numbers, sources, rules, or assumptions — by better using the sources it finds to answer your question."

OpenAI also claims the new GPT-5.6 Sol will produce "more direct responses" with "tighter formatting," and that ChatGPT will "avoid extra detail when it does not help." These are OpenAI's own characterizations; no independent benchmark scores or third-party evaluations were provided alongside the announcement.

Plus and Pro users are also getting a new slider that lets them manually control "how much thought ChatGPT puts into an answer" — effectively a user-facing dial on reasoning depth versus response speed. Both the updated GPT-5.6 Sol and the slider go live Thursday.

What this means

This is a tiering and retention play as much as a model update. Unlimited text access removes a common friction point for free users who hit walls mid-conversation, likely reducing churn toward competitors like Google's Gemini or Anthropic's Claude, both of which offer generous free tiers. Giving free users a "Think" button also narrows the reasoning gap between free and paid tiers, at least for occasional use, while OpenAI presumably still gates heavy reasoning usage through rate limits on non-text inputs.

For paying users, the GPT-5.6 Sol changes target one of the most persistent complaints about large language models: factual reliability tied to dates, numbers, and sourcing. OpenAI's own description — "fewer mistakes" rather than a claim of solved hallucination — is notably restrained, and without published benchmark deltas (on something like SimpleQA or a factuality-specific eval), it's not possible to verify the magnitude of improvement. The addition of a manual reasoning-depth slider suggests OpenAI is moving away from fully automatic model routing toward giving users more direct control over the latency/quality tradeoff, a pattern also seen in Anthropic's extended thinking controls for Claude.