OpenAI is removing the clock, for now
OpenAI is temporarily lifting one of the most visible constraints around its work-focused AI tools: the five-hour usage limit.
Digital Trends reports that the change applies to Plus, Business, and Pro users, and covers both ChatGPT Work and Codex. The company is also making GPT-5.6 Sol more efficient, which should help users get more out of the same level of usage.
That “for now” matters. This is not OpenAI declaring that heavy AI work is suddenly free of limits forever. It is a test of what happens when advanced users stop rationing their sessions and start treating the tools as something that can stay open while the work is actually happening.
The official framing already points in that direction. OpenAI describes ChatGPT Work as built for “your most ambitious work,” while its Codex update for teams focuses on flexible pricing. That combination says a lot. OpenAI is not just selling a faster chatbot response. It is pushing ChatGPT and Codex closer to the center of the workday.
Usage caps interrupt that story. A five-hour ceiling makes sense when AI is treated like a premium compute meter. It makes less sense when a developer wants Codex to stay with a project, or when a team wants ChatGPT Work to sit across drafts, files, meetings, and revisions. Longer sessions change the behavior because they reduce the need to ask whether a task is “worth” spending AI time on.
AI work is becoming harder to meter
This is the bigger shift underneath the update. The more AI tools move from quick prompts into ongoing work sessions, the more the old usage model starts to feel awkward.
People do not work in neat five-hour blocks of AI need. They move between tasks, come back to decisions, ask for edits, test alternatives, and hand off work. Codex is especially exposed to that reality because coding is rarely a single prompt-and-answer exchange. It is iteration, context, debugging, review, and cleanup. If the assistant disappears because a limit has been hit, the product stops feeling like a collaborator and starts feeling like a meter running in the background.
For OpenAI, temporarily loosening limits also gives it a cleaner read on demand. When users are less constrained, the company can see which workflows actually expand, which teams lean hardest on Codex, and whether more efficient models like GPT-5.6 Sol can ease the pressure without making the product feel restricted.
For businesses, the question is no longer simply whether employees should use AI. Many already do. The more practical question is how much of the workday should be designed around AI assistance, and who controls the cost, access, and governance when that assistance becomes persistent.
That is why this small limit change matters. It nudges ChatGPT Work and Codex away from occasional productivity boosts and toward something more embedded. The strategic consequence is clear: in workplace AI, the winning product may not be the one that answers fastest, but the one that can stay with the work longest without making users feel the meter.