my custom gpt truncates file contents
#hi
faint flower
fiery willowBOT
Hi @faint flower. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:
faint flower
my custom gpt truncates file contents
fiery willowBOT
Your description matches expected behavior: when inputs (including file contents) get large, OpenAI systems truncate to stay within the model’s context window.
From the docs:
- For the Responses API, you can control this with the
truncationparameter:auto: if the input exceeds the model’s context window, older items are dropped from the beginning.disabled(default): if the input would exceed the context window, the request fails with a 400 error instead of truncating. [Responses API]
- The same
truncationsetting is exposed in theModelSettings.truncationfield in theopenai-agents-pythonlibrary. [Model settings]
For ChatGPT / custom GPTs specifically, the knowledge sources don’t document a direct setting you can change inside the ChatGPT UI to disable or customize truncation. I’m sorry, but there’s no information in these sources about a “no truncation” toggle for custom GPTs themselves.
However, for ChatGPT Enterprise file uploads, there is documented “context stuffing” behavior:
- Up to ~110k tokens from uploaded documents can be directly included in the model’s context.
- If a single document exceeds 110k tokens, only the first 110k tokens are included; the rest goes only into a private search index (vector store). [Optimizing file uploads]
So if your custom GPT is:
- Running in Enterprise and you upload a very large file, it’s expected that only the first ~110k tokens are “stuffed” into context and the rest is only retrievable via search, which can feel like truncation of the file contents.
If you can share:
- Whether you’re using ChatGPT Enterprise or the API, and
- Rough size of the file (pages or MB),