ChatGPT Drops Message Limits, and Free Users Drop a Model Tier

Anyone who has used ChatGPT's free tier for real work knows the rhythm: you hit the cap around midday, wait it out, and pick up where you left off in the evening. On August 6, OpenAI removed that cap for text conversations. Every user, free accounts included, can now type as much as they want.

The second half of the announcement has drawn far less attention: the model sitting behind the free tier changed, and it got smaller.

Who gets what now

Free and Go users default to GPT-5.6 Luna, replacing GPT-5.5, alongside a new "Think" button that lets the model spend longer on harder questions. Plus and Pro subscribers get an upgraded GPT-5.6 Sol plus a slider that controls how detailed a response should be.

The unlimited part applies to text only. Caps on file uploads, image generation and voice conversations all remain in place. And there is a quieter consequence: free accounts no longer reach OpenAI's frontier reasoning models at all. Unlimited is not the same as unrestricted.

OpenAI says the new Luna produces 62 percent fewer factual errors than its predecessor, and Sol 68 percent fewer. Both figures come from the company's own internal evaluations, with no independent verification published. Worth knowing, not worth treating as a guarantee.

Reading the timing

None of this is happening in isolation. ChatGPT recently passed a billion weekly active users, and competition has shifted decisively onto price. On July 30, OpenAI cut Luna's API pricing by 80 percent: a million input tokens went from one dollar to twenty cents, and a million output tokens from six dollars to $1.20.

Put the two moves side by side and the strategy reads clearly. The cheap, efficient small model is doing double duty, absorbing free-tier demand while undercutting rivals in the developer market. The larger model stays behind the subscription wall as the thing you actually pay for.

How a small business should read it

For everyday work this is straightforwardly good. Drafting emails, writing product descriptions, summarising meeting notes, tightening up a paragraph: the gap between a small model and a large one rarely shows on tasks like these, while losing the daily cap saves real time.

The gap appears when the work gets heavier. Finding the risky clause in a long contract, reconciling financial statements that disagree with each other, extracting requirements from a dense technical spec: reasoning capacity decides the outcome. If you are pushing that kind of task through a free account, check the output more carefully than you used to.

A simple split works well in practice. The free tier is fine for the whole team, but the one or two people handling contracts, finance and technical analysis should sit on a paid plan. That monthly difference is trivial next to the cost of a single clause nobody caught.

The API price cut carries its own consequence. If you shelved a document-processing pipeline or a custom assistant built on your own data because the running cost did not add up, the arithmetic has changed materially. At a fifth of the per-token cost, plenty of scenarios that were filed under "too expensive" deserve another look.

Sources: TechCrunch, The Decoder, OpenAI