mirror of
https://github.com/open-webui/open-webui.git
synced 2026-08-30 17:25:30 -05:00
[GH-ISSUE #24448] issue: Bundled open-webui:ollama image is pinned to Ollama 0.21.2, blocking nemotron3 which requires Ollama 0.22.0 #74910
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: github-starred/open-webui#74910
Reference in New Issue
Block a user
Originally created by @rob-hendrickson on GitHub (May 8, 2026).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/24448
Check Existing Issues
Installation Method
Docker
Open WebUI Version
v0.9.2 latest
Ollama Version (if applicable)
0.21.2
Operating System
Ubuntu 24.04.4 LTS; DGX_OTA_VERSION="7.5.0"
Browser (if applicable)
multiple: safari, chrome, firefox
Confirmation
README.md.Expected Behavior
The bundled image ghcr.io/open-webui/open-webui:ollama should be able to pull and use current Ollama library models such as nemotron3, or the bundled Ollama version should be updated when a newer runtime is required.
Actual Behavior
The bundled image uses Ollama 0.21.2. When attempting to pull nemotron3, the UI returns a manifest 412 error stating that the model requires a newer version of Ollama, so the model cannot be downloaded or used.
Steps to Reproduce
Hi team — this is a specific compatibility blocker for the bundled image ghcr.io/open-webui/open-webui:ollama.
I have personally confirmed that the bundled image is using Ollama 0.21.2, and when I try to run the new nemotron3 model it fails explicitly because that model requires Ollama 0.22.0. Ollama’s nemotron3 library page is live now, and Ollama also announced that Nemotron 3 Nano Omni requires the Ollama 0.22 release.
I also verified the fix path on my side: after rebuilding Open WebUI with Ollama 0.22.0, nemotron3 loaded successfully. That strongly suggests this is not a model-side issue or a Spark-side issue, but a bundled-image version lag in open-webui:ollama.
Because of that, this looks like a high-impact, low-complexity maintenance update: please rebuild and publish ghcr.io/open-webui/open-webui:ollama with Ollama 0.22.0 or newer. Prior discussion around bundled Ollama support already suggests that updating Ollama inside the container is feasible, so this appears to be a straightforward packaging refresh rather than a large architectural change.
For clarity, my results are following the https://build.nvidia.com/spark/open-webui/instructions:
ghcr.io/open-webui/open-webui:ollama → Ollama 0.21.2 → nemotron3 fails because Ollama 0.21.2 is unsupported.
Rebuilt Open WebUI with Ollama 0.22.0 → nemotron3 loads successfully.
I reverted to the official bundled image because I need the maintained path for the rest of my stack.
Would you please prioritize a refresh of the bundled :ollama image? This seems like the kind of small version bump that could unblock users very quickly.
Logs & Screenshots
server:~$ echo "== bundled ollama version ==" && docker exec -it open-webui ollama --version
echo
echo "== live pull attempt ==" && docker logs -f open-webui 2>&1 | grep --line-buffered -E 'ollama/api/pull|apipull|manifest|412|newer version'
== bundled ollama version ==
ollama version is 0.21.2
== live pull attempt ==
2026-05-08 05:17:00.534 | INFO | uvicorn.protocols.http.httptools_impl:send:483 - 172.18.0.1:62016 - "GET /manifest.json HTTP/1.1" 200
2026-05-08 05:17:29.858 | INFO | uvicorn.protocols.http.httptools_impl:send:483 - 172.18.0.1:62026 - "POST /ollama/api/pull/0 HTTP/1.1" 200
2026-05-08 05:17:38.997 | INFO | uvicorn.protocols.http.httptools_impl:send:483 - 172.18.0.1:62027 - "POST /ollama/api/pull/0 HTTP/1.1" 200
2026-05-08 05:18:03.789 | INFO | uvicorn.protocols.http.httptools_impl:send:483 - 172.18.0.1:62030 - "POST /ollama/api/pull/0 HTTP/1.1" 200
2026-05-08 05:18:49.907 | INFO | uvicorn.protocols.http.httptools_impl:send:483 - 172.18.0.1:62032 - "POST /ollama/api/pull/0 HTTP/1.1" 200
2026-05-08 05:23:38.242 | INFO | uvicorn.protocols.http.httptools_impl:send:483 - 172.18.0.1:62048 - "POST /ollama/api/pull/0 HTTP/1.1" 200
2026-05-08 05:23:47.494 | INFO | uvicorn.protocols.http.httptools_impl:send:483 - 172.18.0.1:62049 - "POST /ollama/api/pull/0 HTTP/1.1" 200
2026-05-08 14:48:56.593 | INFO | uvicorn.protocols.http.httptools_impl:send:483 - 172.18.0.1:63353 - "POST /ollama/api/pull/0 HTTP/1.1" 200
2026-05-08 15:26:23.608 | INFO | uvicorn.protocols.http.httptools_impl:send:483 - 172.18.0.1:63444 - "POST /ollama/api/pull/0 HTTP/1.1" 200
Additional Information
TL DR;
NVIDIA dgx spark cannot run the new NVIDIA nemotron 3 Omni using the NVIDIA endorsed recipe on https://build.NVIDIA.com/spark/open-webui/ instructions because ghcr.io/open-webui/open-webui:OLLAMA is compiled with outdated OLLAMA 0.21.2. Recompiling with OLLAMA 0.22.0 fixes this immediately and verifiably.
Please recompile ghcr.io/open-webui/open-webui:ollama with Ollama 0.22.0
Thank you!
@owui-terminator[bot] commented on GitHub (May 8, 2026):
🔍 Related Issues Found
I found some existing issues that might be related. Please check if any of these are duplicates or contain helpful solutions:
🟣 #23886 issue: Ollama Version Out of Date in ghcr.io/open-webui/open-webui:ollama
This is the closest prior report: it is specifically about the bundled
ghcr.io/open-webui/open-webui:ollamaimage shipping an outdated Ollama version and asks for a rebuild to update it. Your issue is the same class of problem, just for a newer Ollama release requirement.by au70ma70n ·
bug🟣 #16306 release new version with bundled ollama 0.11 to support gpt-oss models
This closed issue shows the same pattern of users needing a newer bundled Ollama in Open WebUI to support newly released models. It is relevant as a precedent for the packaging/update request, even though it targets an older Ollama bump and a different model family.
by sbe-arg ·
bug💡 If your issue is a duplicate, please close it and add any additional details to the existing issue instead.
This comment was generated automatically. React with 👍 if helpful, 👎 if not.
@tjbck commented on GitHub (May 8, 2026):
Will be addressed in the next release.