github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #13227] Ollama evicts previously loaded model, although system memory is suficient #8745

New Issue
Closed
opened 2026-04-12 21:30:52 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-12 21:30:52 -05:00
Owner
Copy Link

Originally created by @ioannis-soukas on GitHub (Nov 24, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/13227

Originally assigned to: @jessegross on GitHub.

What is the issue?

Hello,

I am running Ollama v0.13.0 CPU only, on a dedicated ubuntu server with 512 GB RAM.

Although in previous versions I could keep multiple models loaded in memory (Environment="OLLAMA_KEEP_ALIVE=-1"), after the last update it only keeps one model in memory, even when using only small models. Model eviction happens as soon as I load another model.

For example, it can not have gemma3:1b & gemma3:27b loaded at the same time, although the total ram they consume is small, compared to my free RAM. The server is only used for ollama & open-webui, there is no other software running or consuming RAM.

Thank you!

Relevant log output

Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=sched.go:443 msg="system memory" total="503.6 GiB" free="498.1 GiB" free_swap="8.0 GiB"
Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=server.go:702 msg="loading model" "model layers"=63 requested=-1
Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=server.go:974 msg="model requires more memory than is currently available, evicting a model to make space" "loaded layers"=0
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.022Z level=INFO source=runner.go:1398 msg="starting ollama engine"
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.022Z level=INFO source=runner.go:1433 msg="Server listening on 127.0.0.1:41401"
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.352Z level=INFO source=sched.go:443 msg="system memory" total="503.6 GiB" free="499.3 GiB" free_swap="8.0 GiB"
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.352Z level=INFO source=server.go:702 msg="loading model" "model layers"=63 requested=-1
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.353Z level=INFO source=runner.go:1271 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:true KvSize:4096 KvCacheType: NumThreads:28 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.454Z level=INFO source=ggml.go:136 msg="" architecture=gemma3 file_type=Q4_0 name="" description="" num_tensors=1247 num_key_values=40

OS

Ubuntu 24.04.3 LTS

GPU

None

CPU

2xIntel E5-2680v4

Ollama version

0.13.0

Originally created by @ioannis-soukas on GitHub (Nov 24, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/13227 Originally assigned to: @jessegross on GitHub. ### What is the issue? Hello, I am running Ollama v0.13.0 CPU only, on a dedicated ubuntu server with 512 GB RAM. Although in previous versions I could keep multiple models loaded in memory (Environment="OLLAMA_KEEP_ALIVE=-1"), after the last update it only keeps one model in memory, even when using only small models. Model eviction happens as soon as I load another model. For example, it can not have gemma3:1b & gemma3:27b loaded at the same time, although the total ram they consume is small, compared to my free RAM. The server is only used for ollama & open-webui, there is no other software running or consuming RAM. Thank you! ### Relevant log output ```shell Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=sched.go:443 msg="system memory" total="503.6 GiB" free="498.1 GiB" free_swap="8.0 GiB" Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=server.go:702 msg="loading model" "model layers"=63 requested=-1 Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=server.go:974 msg="model requires more memory than is currently available, evicting a model to make space" "loaded layers"=0 Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.022Z level=INFO source=runner.go:1398 msg="starting ollama engine" Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.022Z level=INFO source=runner.go:1433 msg="Server listening on 127.0.0.1:41401" Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.352Z level=INFO source=sched.go:443 msg="system memory" total="503.6 GiB" free="499.3 GiB" free_swap="8.0 GiB" Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.352Z level=INFO source=server.go:702 msg="loading model" "model layers"=63 requested=-1 Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.353Z level=INFO source=runner.go:1271 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:true KvSize:4096 KvCacheType: NumThreads:28 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.454Z level=INFO source=ggml.go:136 msg="" architecture=gemma3 file_type=Q4_0 name="" description="" num_tensors=1247 num_key_values=40 ``` ### OS Ubuntu 24.04.3 LTS ### GPU None ### CPU 2xIntel E5-2680v4 ### Ollama version 0.13.0
GiteaMirror added the bug label 2026-04-12 21:30:52 -05:00
GiteaMirror commented 2026-04-12 21:30:54 -05:00
Author
Owner
Copy Link

@Bottlecap202 commented on GitHub (Nov 24, 2025):

Ask GitHub copilot using grok.

On Mon, Nov 24, 2025, 8:22 AM ioannis-soukas @.***>
wrote:

ioannis-soukas created an issue (ollama/ollama#13227)
https://github.com/ollama/ollama/issues/13227
What is the issue?

Hello,

I am running Ollama v0.13.0 CPU only, on a dedicated ubuntu server with
512 GB RAM.

Although in previous versions I could keep multiple models loaded in
memory (Environment="OLLAMA_KEEP_ALIVE=-1"), after the last update it only
keeps one model in memory, even when using only small models. Model
eviction happens as soon as I load another model.

For example, it can not have gemma3:1b & gemma3:27b loaded at the same
time, although the total ram they consume is small, compared to my free
RAM. The server is only used for ollama & open-webui, there is no other
software running or consuming RAM.

Thank you!
Relevant log output

Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=sched.go:443 msg="system memory" total="503.6 GiB" free="498.1 GiB" free_swap="8.0 GiB"
Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=server.go:702 msg="loading model" "model layers"=63 requested=-1
Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=server.go:974 msg="model requires more memory than is currently available, evicting a model to make space" "loaded layers"=0
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.022Z level=INFO source=runner.go:1398 msg="starting ollama engine"
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.022Z level=INFO source=runner.go:1433 msg="Server listening on 127.0.0.1:41401"
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.352Z level=INFO source=sched.go:443 msg="system memory" total="503.6 GiB" free="499.3 GiB" free_swap="8.0 GiB"
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.352Z level=INFO source=server.go:702 msg="loading model" "model layers"=63 requested=-1
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.353Z level=INFO source=runner.go:1271 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:true KvSize:4096 KvCacheType: NumThreads:28 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.454Z level=INFO source=ggml.go:136 msg="" architecture=gemma3 file_type=Q4_0 name="" description="" num_tensors=1247 num_key_values=40

OS

No response
GPU

No response
CPU

No response
Ollama version

No response

—
Reply to this email directly, view it on GitHub
https://github.com/ollama/ollama/issues/13227, or unsubscribe
https://github.com/notifications/unsubscribe-auth/BDHQPUNBXD5B6KZXQYOP7CT36MICLAVCNFSM6AAAAACNBAIQ76VHI2DSMVQWIX3LMV43ASLTON2WKOZTGY2TSMJTGIYDANI
.
You are receiving this because you are subscribed to this thread.Message
ID: @.***>

<!-- gh-comment-id:3571042516 --> @Bottlecap202 commented on GitHub (Nov 24, 2025): Ask GitHub copilot using grok. On Mon, Nov 24, 2025, 8:22 AM ioannis-soukas ***@***.***> wrote: > *ioannis-soukas* created an issue (ollama/ollama#13227) > <https://github.com/ollama/ollama/issues/13227> > What is the issue? > > Hello, > > I am running Ollama v0.13.0 CPU only, on a dedicated ubuntu server with > 512 GB RAM. > > Although in previous versions I could keep multiple models loaded in > memory (Environment="OLLAMA_KEEP_ALIVE=-1"), after the last update it only > keeps one model in memory, even when using only small models. Model > eviction happens as soon as I load another model. > > For example, it can not have gemma3:1b & gemma3:27b loaded at the same > time, although the total ram they consume is small, compared to my free > RAM. The server is only used for ollama & open-webui, there is no other > software running or consuming RAM. > > Thank you! > Relevant log output > > Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=sched.go:443 msg="system memory" total="503.6 GiB" free="498.1 GiB" free_swap="8.0 GiB" > Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=server.go:702 msg="loading model" "model layers"=63 requested=-1 > Nov 24 13:55:06 llm ollama[7189]: time=2025-11-24T13:55:06.999Z level=INFO source=server.go:974 msg="model requires more memory than is currently available, evicting a model to make space" "loaded layers"=0 > Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.022Z level=INFO source=runner.go:1398 msg="starting ollama engine" > Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.022Z level=INFO source=runner.go:1433 msg="Server listening on 127.0.0.1:41401" > Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.352Z level=INFO source=sched.go:443 msg="system memory" total="503.6 GiB" free="499.3 GiB" free_swap="8.0 GiB" > Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.352Z level=INFO source=server.go:702 msg="loading model" "model layers"=63 requested=-1 > Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.353Z level=INFO source=runner.go:1271 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:true KvSize:4096 KvCacheType: NumThreads:28 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" > Nov 24 13:55:07 llm ollama[7189]: time=2025-11-24T13:55:07.454Z level=INFO source=ggml.go:136 msg="" architecture=gemma3 file_type=Q4_0 name="" description="" num_tensors=1247 num_key_values=40 > > OS > > *No response* > GPU > > *No response* > CPU > > *No response* > Ollama version > > *No response* > > — > Reply to this email directly, view it on GitHub > <https://github.com/ollama/ollama/issues/13227>, or unsubscribe > <https://github.com/notifications/unsubscribe-auth/BDHQPUNBXD5B6KZXQYOP7CT36MICLAVCNFSM6AAAAACNBAIQ76VHI2DSMVQWIX3LMV43ASLTON2WKOZTGY2TSMJTGIYDANI> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> >
GiteaMirror commented 2026-04-12 21:30:54 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 24, 2025):

A full log with OLLAMA_DEBUG=1 in the server environment might help in debugging.

<!-- gh-comment-id:3571081738 --> @rick-github commented on GitHub (Nov 24, 2025): A full log with `OLLAMA_DEBUG=1` in the server environment might help in debugging.
GiteaMirror commented 2026-04-12 21:30:55 -05:00
Author
Owner
Copy Link

@ioannis-soukas commented on GitHub (Nov 24, 2025):

ollama_debug.txt

<!-- gh-comment-id:3571163734 --> @ioannis-soukas commented on GitHub (Nov 24, 2025): [ollama_debug.txt](https://github.com/user-attachments/files/23725448/ollama_debug.txt)
GiteaMirror commented 2026-04-12 21:30:56 -05:00
Author
Owner
Copy Link

@ioannis-soukas commented on GitHub (Nov 24, 2025):

Tried to downgrade a couple of versions & see when this issue first appears.
It seems that 0.12.6 works fine.
The behavior that I described above first appears in v0.12.7.

<!-- gh-comment-id:3572798207 --> @ioannis-soukas commented on GitHub (Nov 24, 2025): Tried to downgrade a couple of versions & see when this issue first appears. It seems that 0.12.6 works fine. The behavior that I described above first appears in v0.12.7.
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#8745
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?