Talking to ChatGPT used to come with a hidden meter running. Just when a conversation got going, the limit could stop you cold.
That is ending. OpenAI announced on August 6, 2026 that text chats will no longer be capped anywhere on ChatGPT, a service that has just passed one billion people using it each week.
The rollout is staggered. Plus and Pro subscribers get their upgrade today, while Free and Go accounts start getting changes this week and unlimited text chats plus the Think button arrive next week.
For the people using ChatGPT to think things through, the change removes a familiar frustration. You no longer have to wonder whether the next question will be the one that hits the wall.
OpenAI is splitting its lineup into two paths.
The new GPT-5.6 Luna model will be the default for Free and Go users, replacing GPT-5.5. Luna is the everyday model, the one you use for quick answers and casual conversation.
The split makes sense. A quick question does not need the same horsepower as a deep planning session, so giving each group a model tuned for its own job keeps things fast for everyone.
For harder questions, there is a Think button. Press it when a question needs extra reasoning, and the model takes a beat to work it through.
Get the free Always Be Buying eBook and learn the simple system for building wealth on any income
The enhanced GPT-5.6 Sol for Plus and Pro subscribers is built for quick jobs, including questions, planning, writing, web research, advice, and decisions. OpenAI says the responses are more concise and hold up better under pressure.
There is also a thinking slider for Plus and Pro users. It sets how much reasoning the model uses, so a simple question stays quick and a multi-step one gets room to breathe.
One distinction worth knowing: the upgraded Sol is not the same as the GPT-5.6 Sol used in Codex and Work. That version stays unchanged.
Caps are not disappearing everywhere. Image generation, images, voice, and file uploads keep their own separate limits, and only text chats are becoming unlimited.
Internal testing gives a sense of how much sharper these models are. GPT-5.6 Luna made 62% fewer factual errors than GPT-5.5-Instant in OpenAI's internal testing.
For GPT-5.6 Sol, factual errors were 68% less common in an internal evaluation. These are OpenAI's own numbers, not independent checks.
Fewer factual errors is another way of saying the models get things wrong less often. For anyone who has watched a chatbot sound completely sure of itself and still be wrong, that is a meaningful upgrade.
Even with the improvement, these are still machines that can sound confident and be wrong. The lower error counts are progress, not perfection.
The business angle here goes beyond a new feature. It is a signal about how OpenAI plans to win.
OpenAI could have kept limits in place and charged extra for more use. Instead, it is giving a billion people free text chats, which is a bet that scale beats short-term revenue.
That bet sends pressure across the AI industry. The biggest player is raising quality and making the product easier to use at the same time, and the rest of the market has to answer on both fronts.
Over time, that pressure lands in the tech stocks you own. The companies that can keep up with better models and wider access are the ones best positioned for the next stage of the race.
The race to watch is who can match both moves. When the leader improves the product and lowers the cost of entry in one announcement, the follow-up moves from rivals are usually where the market shows who is really competing.
Download the free Always Be Buying eBook and start putting your money to work today
