[GH-ISSUE #3515] [BUG]: High cpu usage #2267

Closed
opened 2026-02-22 18:28:54 -05:00 by yindo · 7 comments
Owner

Originally created by @tanuki-cat on GitHub (Mar 22, 2025).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/3515

How are you running AnythingLLM?

AnythingLLM desktop app

What happened?

When using the DeepSeek R1 model, the CPU usage is high when waiting for a response or rendering, almost using up all idle resources. Is it really necessary to consume such high resources when waiting for a response or rendering? Is this an unpredictable bug or is it software design?
Hope to get a reply. Thanks

Image

Are there known steps to reproduce?

No response

Originally created by @tanuki-cat on GitHub (Mar 22, 2025). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/3515 ### How are you running AnythingLLM? AnythingLLM desktop app ### What happened? When using the DeepSeek R1 model, the CPU usage is high when waiting for a response or rendering, almost using up all idle resources. Is it really necessary to consume such high resources when waiting for a response or rendering? Is this an unpredictable bug or is it software design? Hope to get a reply. Thanks ![Image](https://github.com/user-attachments/assets/75342ef3-01b9-43ed-bbf1-24162af3f585) ### Are there known steps to reproduce? _No response_
yindo added the possible bugDesktopinvestigating labels 2026-02-22 18:28:54 -05:00
yindo closed this issue 2026-02-22 18:28:54 -05:00
Author
Owner

@timothycarambat commented on GitHub (Mar 24, 2025):

To be clear, you are not using the built in LLM are you? If you dont have a GPU and are running models on CPU some systems show the Renderer process as high CPU when really it is the main thread doing LLM Inference on CPU.

Outside of that, what are your system specs (CPU, RAM, OS) and what version of AnythingLLM Desktop are you on?

@timothycarambat commented on GitHub (Mar 24, 2025): To be clear, you are _not_ using the built in LLM are you? If you dont have a GPU and are running models on CPU some systems show the Renderer process as high CPU when really it is the main thread doing LLM Inference on CPU. Outside of that, what are your system specs (CPU, RAM, OS) and what version of AnythingLLM Desktop are you on?
Author
Owner

@tanuki-cat commented on GitHub (Mar 26, 2025):

I'm not running any mods on the cpu, using a call to the api interface.
My system related parameters are as follows:
Cpu:M1Max 10cpu 32gpu
System OS: macos 15.3.2 (24D81)
Memory RAM: 32GB

And AnythingLLM Despktop Version: 1.7.7

@tanuki-cat commented on GitHub (Mar 26, 2025): I'm not running any mods on the cpu, using a call to the api interface. My system related parameters are as follows: Cpu:M1Max 10cpu 32gpu System OS: macos 15.3.2 (24D81) Memory RAM: 32GB And AnythingLLM Despktop Version: 1.7.7
Author
Owner

@jarkko-hautakorpi commented on GitHub (Jun 18, 2025):

Same here. In Ubuntu the AnythingLLM desktop uses constantly 40% CPU even when application has no activity in long time. I only use LLM's via API, no locals. Inactive ubuntu desktop app eating 40% CPU all the time, when it should be idle. I need to close the program when I don't use it to save CPU, otherwise my Lenovo laptop i7 23Gb RAM becomes too sluggish.

@jarkko-hautakorpi commented on GitHub (Jun 18, 2025): Same here. In Ubuntu the AnythingLLM desktop uses constantly 40% CPU even when application has no activity in long time. I only use LLM's via API, no locals. Inactive ubuntu desktop app eating 40% CPU all the time, when it should be idle. I need to close the program when I don't use it to save CPU, otherwise my Lenovo laptop i7 23Gb RAM becomes too sluggish.
Author
Owner

@timothycarambat commented on GitHub (Jun 24, 2025):

@jarkko-hautakorpi When in that state, is the application window totally minimized/hidden, or is it in the UI and just unused? Since all of this comes from the render process its not related to models or anything and almost certainly has to do with animations in the UI, however there should not be any animations in an idle state.

@timothycarambat commented on GitHub (Jun 24, 2025): @jarkko-hautakorpi When in that state, is the application window totally minimized/hidden, or is it in the UI and just unused? Since all of this comes from the render process its not related to models or anything and almost certainly has to do with animations in the UI, however there should not be any animations in an idle state.
Author
Owner

@avinashkanaujiya commented on GitHub (Jul 10, 2025):

I'm also facing a similar issue. When it's not even in the foreground and I'm using some other app, it constantly keeps using CPU cycles in the background, as if it's doing something, but in reality, it's not.

I don't use any local models. I use all the APIs, so there's no question for it to be using that many CPU cycles.

@avinashkanaujiya commented on GitHub (Jul 10, 2025): I'm also facing a similar issue. When it's not even in the foreground and I'm using some other app, it constantly keeps using CPU cycles in the background, as if it's doing something, but in reality, it's not. I don't use any local models. I use all the APIs, so there's no question for it to be using that many CPU cycles.
Author
Owner

@Peppe289 commented on GitHub (Oct 23, 2025):

I’m facing the same issue on Arch Linux with GNOME. The UI process alone consumes 100% CPU usage whenever the application window is open, even when no LLM or additional workloads are running. The problem does not occur when using KDE.

@Peppe289 commented on GitHub (Oct 23, 2025): I’m facing the same issue on Arch Linux with GNOME. The UI process alone consumes 100% CPU usage whenever the application window is open, even when no LLM or additional workloads are running. The problem does not occur when using KDE.
yindo changed title from [BUG]: High cpu usage to [GH-ISSUE #3515] [BUG]: High cpu usage 2026-06-05 14:45:32 -04:00
Author
Owner

@asileng commented on GitHub (May 20, 2026):

In my case, I noticed that this issue only happens when you have a too long context

<!-- gh-comment-id:4494705858 --> @asileng commented on GitHub (May 20, 2026): In my case, I noticed that this issue only happens when you have a too long context
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#2267