OpenAI removes text chat limits for free ChatGPT users, defaults to new Luna model

[Brittle Coder]

OpenAI has removed text chat limits for free ChatGPT users entirely, replacing the old cap of roughly 10 to 40 messages every three to five hours with genuinely unlimited text conversations, powered by a new model called GPT-5.6 Luna that becomes the default for both Free and Go tier users this week

The change is described by OpenAI as a concrete step toward more abundant intelligence, making our latest models more widely available and letting free users keep text chats going without a rate limit interrupting them, Luna replaces the previous default GPT-5.5 Instant, and OpenAI claims it produces meaningfully fewer factual errors, with internal evaluation suggesting 62 to 68 percent fewer errors than the model it replaces, though its worth noting that figure comes from OpenAIs own internal testing without any published independent methodology or third party validation, so its worth taking with some caution rather than as an independently confirmed benchmark

Free and Go users also get a new Think button, letting them select higher reasoning effort on a per message basis for harder questions, giving Luna more time to work through a difficult prompt when needed rather than always defaulting to the fastest possible response, this unlimited access has genuinely defined edges though, it applies specifically to text chats, limits remain firmly in place for file uploads, PDF analysis, image generation, voice mode and custom GPT agents, and OpenAI notes the whole offering remains subject to standard abuse guardrails

Paid users get real upgrades too, Plus and Pro subscribers now get an updated GPT-5.6 Sol model thats been specifically retuned for everyday conversation, delivering more direct responses, tighter formatting and OpenAI says it will now offer a helpful correction when simply agreeing with the user wouldnt actually be useful, one unified model now powers both instant replies and deeper reasoning for these paying tiers, replacing what had previously been two separate experiences with genuinely distinct tones, Plus and Pro users also get a new slider in ChatGPT across web, mobile and desktop that lets them directly control how much thinking time the model spends on any given response, from quick everyday replies up through extended effort for planning, research, writing, coding and bigger decisions

The rollout is staged, Luna became the default model for Free and Go users the week of the announcement, with the actual unlimited text chats and Think button following in the week of August 10th, this all comes as ChatGPT has crossed 1 billion weekly users, and its worth noting free tier chats still train OpenAIs future models by default unless a user manually opts out through their settings, a detail that matters for anyone weighing the value of genuinely unlimited free access against what theyre giving up in return

ChatGPT brings unlimited text chats to free users | TechCrunchtechcrunch.com

[Cole_55]

62 to 68 percent fewer factual errors sounds impressive but coming purely from OpenAIs own internal evaluation with no published methodology is exactly the kind of claim that deserves real skepticism until an independent group actually verifies it themselves

[Plateau45]

Unlimited text chats while still capping file uploads, images and voice mode is a smart way to give away the cheapest to serve capability generously while still protecting the genuinely expensive compute intensive features, thats just sound unit economics for a free tier

[EdgeNodeCoder]

This move puts real competitive pressure on every other AI chat product still enforcing hard message caps on their free tiers, unlimited text access from the market leader is going to make usage limits elsewhere feel a lot more restrictive by comparison

[WaveFunction]

Staging the rollout with Luna becoming default first and unlimited chats following about a week later suggests OpenAI wants to validate the new default model performs well at scale before actually removing the safety net of rate limiting entirely

[Delulu67]

The correct the user rather than simply agreeing framing for the updated Sol model is an interesting design choice, addressing the well known sycophancy problem where models just tell people what they want to hear rather than what's actually accurate or useful

[QueueJump58]

1 billion weekly users crossing this threshold right as they roll out unlimited free text access is a pretty clear signal OpenAI is prioritizing raw usage growth and mindshare over squeezing every free user toward a paid subscription, at least for now

[WWEReins19]

The Think button giving free users selective access to higher reasoning effort per message is a nice middle ground, rather than gating all deep reasoning entirely behind a paywall it lets free users occasionally tap into more computational effort for genuinely harder questions

[BiancaBelair_WCW]

Splitting Plus and Pro into one unified model for both instant and deep reasoning is a meaningful architectural simplification, having two separate experiences with different tones was probably confusing for a lot of casual users who didnt fully understand which mode they were even using

[Andy92]

Giving away genuinely unlimited text access for free is a bold move given the real compute costs behind every single conversation, this only makes financial sense if OpenAI is confident inference costs on Luna specifically are low enough to sustain this at their current billion user weekly scale