[GH-ISSUE #11054] just downloading manifest counts as a full download #69351

Closed
opened 2026-05-04 17:53:30 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @Sk16er on GitHub (Jun 12, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11054

What is the issue?

So this is a issue which is ollama have a venerability.
which is if any device made a request to download a model, it stats that it downloaded.
meaning let suppose i made a request to download a model, eg.
Please do not delete my model which i made in this recording
this i made to showcase my class student's.
and do not cut the downloads in that.

hope you fix this other wise some could mis use this

Relevant log output


OS

No response

GPU

No response

CPU

No response

Ollama version

No response

Originally created by @Sk16er on GitHub (Jun 12, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11054 ### What is the issue? So this is a issue which is ollama have a venerability. which is if any device made a request to download a model, it stats that it downloaded. meaning let suppose i made a request to download a model, eg. *Please do not delete my model which i made in this recording* this i made to showcase my class student's. and do not cut the downloads in that. hope you fix this other wise some could mis use this ### Relevant log output ```shell ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version _No response_
GiteaMirror added the bugollama.com labels 2026-05-04 17:53:31 -05:00
Author
Owner

@pdevine commented on GitHub (Jun 16, 2025):

This is expected behaviour. Models are broken up into "image layers" which can be shared with other models (and can be referenced by the manifest), so it's possible that a person has already downloaded part of a given model and they don't need to download it again. We also hand off downloads to cloudflare for the individual image layers so it's tricky for ollama.com to know exactly when a download is finished.

The way this gets protected against is through rate limiting. The automatic rate limiting will kick in if you try to hit the API too much, and then at some point we'll just ban the IP address if we see someone abusing it.

<!-- gh-comment-id:2978451125 --> @pdevine commented on GitHub (Jun 16, 2025): This is expected behaviour. Models are broken up into "image layers" which can be shared with other models (and can be referenced by the manifest), so it's possible that a person has already downloaded part of a given model and they don't need to download it again. We also hand off downloads to cloudflare for the individual image layers so it's tricky for ollama.com to know exactly when a download is finished. The way this gets protected against is through rate limiting. The automatic rate limiting will kick in if you try to hit the API too much, and then at some point we'll just ban the IP address if we see someone abusing it.
Author
Owner

@Sk16er commented on GitHub (Jun 17, 2025):

Thanks for response but sorry i don't write the issue correctly,😥. But as you say that if someone abuse this you can block that i donot know that, if this is the case then it is clear

<!-- gh-comment-id:2979334736 --> @Sk16er commented on GitHub (Jun 17, 2025): Thanks for response but sorry i don't write the issue correctly,😥. But as you say that if someone abuse this you can block that i donot know that, if this is the case then it is clear
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#69351