github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-05 23:53:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #11864] Slow performance when web search is enabled in GPT_OSS #54388

New Issue
Closed
opened 2026-04-29 05:52:28 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-29 05:52:28 -05:00
Owner
Copy Link

Originally created by @sakana-max on GitHub (Aug 12, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11864

Originally assigned to: @ParthSareen on GitHub.

What is the issue?

Slow performance when web search is enabled in GPT_OSS

When web search is enabled in GPT_OSS, it uses a lot of CPU power, resulting in very long output times.

The CPU is an i5-12400.
The GPU is an RTX4060TI with 8GB of Vram.
The RAM and internet connection are also sufficient.
The model is gpt_oss:20b.
It seems that the web search itself is slow, and the text never loads, even after waiting.

I used machine translation.

Image Image

Relevant log output

OS

Windows

GPU

Nvidia

CPU

Intel

Ollama version

0.11.4

Originally created by @sakana-max on GitHub (Aug 12, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11864 Originally assigned to: @ParthSareen on GitHub. ### What is the issue? Slow performance when web search is enabled in GPT_OSS When web search is enabled in GPT_OSS, it uses a lot of CPU power, resulting in very long output times. The CPU is an i5-12400. The GPU is an RTX4060TI with 8GB of Vram. The RAM and internet connection are also sufficient. The model is gpt_oss:20b. It seems that the web search itself is slow, and the text never loads, even after waiting. I used machine translation. <img width="1919" height="1026" alt="Image" src="https://github.com/user-attachments/assets/36fe3110-b550-4f28-8fe7-ea11d3a5be76" /> <img width="1919" height="1025" alt="Image" src="https://github.com/user-attachments/assets/5066388b-8a5e-41ec-b710-8838bc6543ec" /> ### Relevant log output ```shell ``` ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version 0.11.4
GiteaMirror added the bug label 2026-04-29 05:52:28 -05:00
GiteaMirror commented 2026-04-29 05:52:35 -05:00
Author
Owner
Copy Link

@ParthSareen commented on GitHub (Aug 12, 2025):

Hey @sakana-max. The gpt-oss model needs at least 16gb vram on the low end so my gut says that it's CPU offloading causing the spike and slowness. Could you show me the output while the model is loaded with ollama ps

<!-- gh-comment-id:3181224868 --> @ParthSareen commented on GitHub (Aug 12, 2025): Hey @sakana-max. The gpt-oss model needs at least 16gb vram on the low end so my gut says that it's CPU offloading causing the spike and slowness. Could you show me the output while the model is loaded with `ollama ps`
GiteaMirror commented 2026-04-29 05:52:37 -05:00
Author
Owner
Copy Link

@sakana-max commented on GitHub (Aug 12, 2025):

Thank you for your reply. Is the output from PS correct?

NAME ID SIZE PROCESSOR CONTEXT UNTIL
gpt-oss:20b e95023cf3b7b 14 GB 54%/46% CPU/GPU 4096 4 minutes from now

I have one more question. When outputting, the same internet search is performed multiple times. Is this correct? Or is this due to insufficient VRAM?

Image
<!-- gh-comment-id:3181319691 --> @sakana-max commented on GitHub (Aug 12, 2025): Thank you for your reply. Is the output from PS correct? NAME ID SIZE PROCESSOR CONTEXT UNTIL gpt-oss:20b e95023cf3b7b 14 GB 54%/46% CPU/GPU 4096 4 minutes from now I have one more question. When outputting, the same internet search is performed multiple times. Is this correct? Or is this due to insufficient VRAM? <img width="692" height="854" alt="Image" src="https://github.com/user-attachments/assets/ce532b44-301c-4f0f-93b4-337395988b2c" />
GiteaMirror commented 2026-04-29 05:52:38 -05:00
Author
Owner
Copy Link

@ParthSareen commented on GitHub (Aug 12, 2025):

Yeah so the issue is that the model is too big for VRAM - the 54%/46% CPU/GPU 4096 split shows that more than half the model is working on CPU which is why you see the slow down. The model has been trained to make multiple searches so this is expected behavior. Sorry. We should have websearch for smaller models coming soon! You can try out turbo in the meantime if that is of interest.

<!-- gh-comment-id:3181332939 --> @ParthSareen commented on GitHub (Aug 12, 2025): Yeah so the issue is that the model is too big for VRAM - the `54%/46% CPU/GPU 4096` split shows that more than half the model is working on CPU which is why you see the slow down. The model has been trained to make multiple searches so this is expected behavior. Sorry. We should have websearch for smaller models coming soon! You can try out turbo in the meantime if that is of interest.
GiteaMirror commented 2026-04-29 05:52:39 -05:00
Author
Owner
Copy Link

@sakana-max commented on GitHub (Aug 12, 2025):

Thank you for your reply. I understand the problem. I'm a junior high school student and can't upgrade my graphics card, so I'll try a different model.

<!-- gh-comment-id:3181411554 --> @sakana-max commented on GitHub (Aug 12, 2025): Thank you for your reply. I understand the problem. I'm a junior high school student and can't upgrade my graphics card, so I'll try a different model.
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#54388
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?