[GH-ISSUE #12474] num_ctx incorrect description in documentation. #54801

Open
opened 2026-04-29 07:20:43 -05:00 by GiteaMirror · 1 comment
Owner

Originally created by @wmantly on GitHub (Oct 1, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12474

What is the issue?

Several places in the documentation define num_ctx as "Sets the size of the context window used to generate the next token." Is this correct? Does num_ctx not refer to the context window? Is the context window not more like the "total number of tokens the model can handle at once, prompt tokens + generated tokens ( + tool calling tokens, ect)"

If so, this will lead people to not understand why ollama isnt behaving when they overtax the context window.

If the description is accurate, how does it meaningfully differ from num_predict

Relevant log output

None, issue refers to documentation.

OS

No response

GPU

No response

CPU

No response

Ollama version

n/a

Originally created by @wmantly on GitHub (Oct 1, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12474 ### What is the issue? Several places in the documentation define `num_ctx` as "Sets the size of the context window used to **generate the next token**." Is this correct? Does `num_ctx` not refer to the context window? Is the context window not more like the "total number of tokens the model can handle at once, prompt tokens + generated tokens ( + tool calling tokens, ect)" If so, this will lead people to not understand why ollama isnt behaving when they overtax the context window. If the description is accurate, how does it meaningfully differ from `num_predict` ### Relevant log output ```shell None, issue refers to documentation. ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version n/a
GiteaMirror added the documentation label 2026-04-29 07:20:43 -05:00
Author
Owner

@mvanhorn commented on GitHub (Mar 9, 2026):

I've submitted a fix in the PR above. Updated the num_ctx description to clarify it controls the total context window size (prompt + output combined).

<!-- gh-comment-id:4024413455 --> @mvanhorn commented on GitHub (Mar 9, 2026): I've submitted a fix in the PR above. Updated the num_ctx description to clarify it controls the total context window size (prompt + output combined).
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#54801