OpenAI removed the five-hour usage cap on GPT-5.6 Sol, its most powerful model, for all paid subscribers on Plus, Business, and Pro plans. The change is temporary, with no end date announced, and comes alongside a one-time reset of everyone's usage counters.
The tools involved are worth understanding. Codex is an AI that can handle software tasks end to end: writing code, reviewing it, running tests, fixing bugs. You give it a goal and it works independently, sometimes for hours. ChatGPT Work is a newer product that does the same thing for general professional tasks, producing finished documents, spreadsheets, or web pages rather than chat replies. Both tools share a single usage pool, meaning heavy use of either eats into the same allowance.
The five-hour window was the visible edge of that system. Under normal rules, after five hours of combined usage, you hit a wall and wait for the clock to reset. For anyone running longer projects through these tools, that interruption is not a minor inconvenience: it breaks workflows mid-task.
The demand surge that triggered this change is telling. OpenAI says the platform crossed six million active users on Codex and ChatGPT Work. A recent OpenAI research paper found that nearly a quarter of all Codex requests are now for tasks that would take a person more than one hour to complete. The five-hour cap was colliding directly with how people were actually using the product.
OpenAI is also rolling out efficiency improvements to GPT-5.6 Sol that reduce how much of your allowance each task consumes. The exact numbers have not been published yet, but the intent is that the same subscription gets you further before you hit the wall. This is a meaningful change even when the temporary cap removal ends.
For businesses evaluating or already using these tools, the pattern here is not just a usage update. It points to a broader shift in how AI is being used: less as a quick assistant and more as something that runs alongside work for hours at a time. The old model, where usage is rationed in short bursts, fits occasional queries. It fits poorly when an AI is mid-way through a multi-hour analysis or a long document review.
The temporary nature of the change is important. OpenAI has not announced permanent new limits, and the five-hour cap will likely return in some form. The efficiency improvements to the model are the more durable development: they mean your subscription stretches further regardless of what the cap ends up being.
If your team uses ChatGPT for anything beyond short questions, right now is a good time to run larger experiments and see how much heavier usage these tools can absorb before limits become relevant to your workflow. That will help you make a more grounded decision about whether a higher-tier plan or additional credits make sense.