What are the limits of code interpreter?

It seems I am able to upload extremely large documents to code interpreter. Anyone know the limits of what it’s able to use as context? Is it using gpt4-32k when you ask it to analyze data?

Anyone have any novel use cases that they havent shared yet?

It is using the standard 8k gpt-4, so even though you can upload large files, you’ll need a system for breaking them up as needed.