In chat, GPT-5 can use the Python tool to make e.g. charts. The same instruction, when scheduled as a Task, won’t run Python and reports it isn’t available in the Task environment. Web browsing and memory still work.
Is this an intentional limitation or a bug? If intentional, is scheduled Python on the roadmap, and are there recommended workarounds beyond using cron/serverless to call back into chat?
It seems that when remotely executing a command it will not generate the virtual machine required to execute the code. You can try emailing support@openai.com and enquiring if this is intended or not, but my guess is that without any way to pull that data back automatically it would simply time out and the data would be lost.
response from support (the same question as OP):
Hi Matej, I’m an AI support agent—happy to help!
You’re correct: when you use ChatGPT-5 (or GPT-5) in a regular chat, you can run Python code with the code interpreter. However, when you set up a Task (automated, scheduled prompt execution), Python/code execution is not supported in Tasks.
This is intentional. As of now, the Tasks feature doesn’t allow Python code execution or code interpreter access. Tasks are limited to web browsing and memory, but not files, Python, or voice.
There’s no published timeline for enabling code execution in Tasks. If you need to automate code execution, your only workaround is to use an external scheduler (cron, serverless, etc.) that calls the ChatGPT API for “chat” completion with the code interpreter enabled, rather than using Tasks.
If you’d like updates or changes to this functionality, I recommend keeping an eye on the official OpenAI changelog and Help Center.
Let me know if you have more questions!
This response was generated with AI support which can make mistakes.