github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #10546] MoE requires loading full model in memory #53452

New Issue
Closed
opened 2026-04-29 03:15:43 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-29 03:15:43 -05:00
Owner
Copy Link

Originally created by @markokocic on GitHub (May 3, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10546

What is the issue?

Tried loading Qwen3 MoE 30b-a3b on my laptop. Since it’s a MoE with 3b active parameters, I expected that ollama would be able to run in with limited memory. However, I got an error message telling me it requires 20+GB to run, which is the amount of memory required to run a fully loaded dense model.

Relevant log output

OS

Windows

GPU

Other

CPU

Intel

Ollama version

No response

Originally created by @markokocic on GitHub (May 3, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10546 ### What is the issue? Tried loading Qwen3 MoE 30b-a3b on my laptop. Since it’s a MoE with 3b active parameters, I expected that ollama would be able to run in with limited memory. However, I got an error message telling me it requires 20+GB to run, which is the amount of memory required to run a fully loaded dense model. ### Relevant log output ```shell ``` ### OS Windows ### GPU Other ### CPU Intel ### Ollama version _No response_
GiteaMirror added the bug label 2026-04-29 03:15:43 -05:00
GiteaMirror commented 2026-04-29 03:15:45 -05:00
Author
Owner
Copy Link

@Burnarz commented on GitHub (May 3, 2025):

Hi! That’s how a MoE model works.
Even though only 3B parameters are active during inference, all the experts need to be loaded into memory so the model can route your input to the appropriate ones based on the request.

<!-- gh-comment-id:2848641808 --> @Burnarz commented on GitHub (May 3, 2025): Hi! That’s how a MoE model works. Even though only 3B parameters are active during inference, all the experts need to be loaded into memory so the model can route your input to the appropriate ones based on the request.
GiteaMirror commented 2026-04-29 03:15:46 -05:00
Author
Owner
Copy Link

@markokocic commented on GitHub (May 3, 2025):

Thanks

<!-- gh-comment-id:2848648712 --> @markokocic commented on GitHub (May 3, 2025): Thanks
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#53452
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?