You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
+1, BYOK/BYOM without Login is now possible in copilot CLI and soon in the vscode plugin (microsoft/vscode#312207). Please enable this for the eclipse plugin too.
Thank you very much for your patience. Ollama is supported on the lastest GitHub Copilot for eclipse (v 0.21.0). For the best stability and user experience, we also recommend upgrading Ollama to the latest version.
Please give it a try and let us know if you have any feedback.
Or let's say, I should think about using a different local model. 😏 But that's a different story.
What thing I noticed are the names displayed in the Model Management dialog: I get
1x Qwen3.6
8x KL0.099
4x Qwen3.5
If I do ollama ls I get different names. I have actually 1x Qwen3.6 and a bunch of Qwen3.5 models installed, but the names are different, like
Qwen3.6-27B-IQ4_XS:latest
Qwen3.5-27B-Q8_0:latest
Qwen3.5-9B-BF16:latest
etc.
I'm not sure which field Copilot uses here, especially as I couldn't find KL0.099 anywhere on the Ollama API as information.
A nice-to-have for custom models would be, to also show the context size on the information card which opens when hovering over the model. As the context status is already showing the correct context size - at least my cases - I assume it should not be a big problem. But as said: That would also be a nice-to-have.
Should I open issues for both? Especially as this one was already closed.
I also noticed that Copilot run my Qwen3.6 model with a 4K context window, as shown by ollama ps. I'm a little bit confused, as I would have expected a 256K context window. 🤔
And also Copilot shows that the model has a 256K context window, the agent runs out of context after 4K:
I don't have context from a previous turn in this conversation. What were you working on? Please share the task or file you'd like me to continue with.
I tried using ollama run and noticed that 4K in my case is the default of Qwen3.6. Means Copilot is using the default, but assuming the maximum.
I checked it also against the standard qwen3.5:latest model which I downloaded using the Ollama UI. Also here the default context is 4K.
Would it be possible, that Copilot - when talking to Ollama - also sets the context window size? Let me also open an issue for this.
Activity
jb-prismat commentedon Feb 25, 2026
any update? :-)
JAlTad commentedon Apr 2, 2026
+1, interested in the feature
gerrit-hohl commentedon Apr 28, 2026
Just opened accidental an issue #124 for the same. I would also be interested in this feature. 👍
bstoltefuss commentedon May 5, 2026
+1, BYOK/BYOM without Login is now possible in copilot CLI and soon in the vscode plugin (microsoft/vscode#312207). Please enable this for the eclipse plugin too.
attilaberencsi commentedon Jun 14, 2026
+1 , to have OpenAI protocol Compatible (local) LLM Server enablement. Thank You :) !
lprellnagarro commentedon Jun 17, 2026
+1 Interest
ethanyhou commentedon Aug 10, 2026
Hi jb-abap (@jb-prismat), JAlTad, bstoltefuss, Attila Berencsi (@attilaberencsi), and lprellnagarro,
Thank you very much for your patience. Ollama is supported on the lastest GitHub Copilot for eclipse (v 0.21.0). For the best stability and user experience, we also recommend upgrading Ollama to the latest version.
Please give it a try and let us know if you have any feedback.
gerrit-hohl commentedon Aug 10, 2026
Thank you, Ethan. 🙏
I just tested it and it works. 👍
Or let's say, I should think about using a different local model. 😏 But that's a different story.
What thing I noticed are the names displayed in the Model Management dialog: I get
Qwen3.6KL0.099Qwen3.5If I do
ollama lsI get different names. I have actually 1x Qwen3.6 and a bunch of Qwen3.5 models installed, but the names are different, likeQwen3.6-27B-IQ4_XS:latestQwen3.5-27B-Q8_0:latestQwen3.5-9B-BF16:latestI'm not sure which field Copilot uses here, especially as I couldn't find
KL0.099anywhere on the Ollama API as information.A nice-to-have for custom models would be, to also show the context size on the information card which opens when hovering over the model. As the context status is already showing the correct context size - at least my cases - I assume it should not be a big problem. But as said: That would also be a nice-to-have.
Should I open issues for both? Especially as this one was already closed.
gerrit-hohl commentedon Aug 10, 2026
I also noticed that Copilot run my Qwen3.6 model with a 4K context window, as shown by
ollama ps. I'm a little bit confused, as I would have expected a 256K context window. 🤔And also Copilot shows that the model has a 256K context window, the agent runs out of context after 4K:
I tried using
ollama runand noticed that 4K in my case is the default of Qwen3.6. Means Copilot is using the default, but assuming the maximum.I checked it also against the standard
qwen3.5:latestmodel which I downloaded using the Ollama UI. Also here the default context is 4K.Would it be possible, that Copilot - when talking to Ollama - also sets the context window size? Let me also open an issue for this.
gerrit-hohl commentedon Aug 10, 2026
Opened issues #409, #410 and #411.