Starting in December, OpenAI will implement a common limit for ChatGPT and Codex on Plus, Pro, and Business plans. Now, active code work may leave fewer tokens for complex chat queries — and vice versa. Users are already expressing dissatisfaction, as they don’t even know how many queries they will have available.
Starting in December, OpenAI will implement a common limit for ChatGPT and Codex on Plus, Pro, and Business plans. Now, active code work may leave fewer tokens for complex chat queries — and vice versa. Users are already expressing dissatisfaction, as they don’t even know how many queries they will have available.
OpenAI reports on the changes.
OpenAI will count ChatGPT and Codex usage in tokens. How much of the shared quota a particular query uses will depend on the model chosen and the complexity of the task. For example, asking an AI to analyze a large software project or perform complex research can take up significantly more tokens than a simple question.
But regular conversations with ChatGPT in the fast Instant mode will not use up the shared limit. You can check how much resource is left and when it will be restored in the Settings → Usage section. If the tokens run out, you will have to wait for them to be renewed, switch to a more expensive tariff or purchase additional credits for features that support it.
OpenAI assures that most subscribers will not have to wait longer for their limits to be restored. However, the company has not yet provided specific figures, so it is difficult to assess how generous the new quota will be.
On GitHub, in the Codex repository, subscribers are already asking OpenAI not to combine limits. One Plus user explained that the ability to use ChatGPT and Codex separately was an important benefit of the service for him. If it disappears, he may cancel his subscription altogether instead of upgrading to a more expensive plan.
Hacker News users also criticized the new system for being opaque, noting that OpenAI offers to pay for a subscription without explaining exactly how many tokens are included or how the new limits compare to the current ones.
This may be because OpenAI is preparing to bring ChatGPT Work's capabilities directly to regular chat. According to BleepingComputer, testing of the new approach should begin in the coming weeks with a small group of users. So you can simply ask ChatGPT, for example, to create a website, prepare a presentation or perform actions on the Internet, and then monitor the process and adjust the result.
Previously, dev.ua wrote that OpenAI plans to combine ChatGPT, Codex, and a browser into a single application. The company explained that working with several separate tools complicates development and prevents maintaining product quality.
