copilot provider models is so much slower lately

[althafdemiandra]

Question

haven't really used copilot with other than opencode (not using copilot chat)
but lately i feels like copilot models is really slow in my opencode

idk where to ask about this since this is mostlikely copilot problem, but like to know is it happening to all of you guys?

it tooks 3-5 minutes with gemini 3 flash, sonnet 4.5 and gpt 5.2 to answer simple question

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• Absurd amount of cpu and memory usage when doing pah completion #496: Absurd amount of cpu and memory usage when doing pah completion
• [FEATURE]: Use a cheap model to name sessions #4133: Use a cheap model to name sessions (addresses expensive model slowdown)
• Poor performance with local models #4385: Poor performance with local models

Feel free to ignore if none of these address your specific case.

[kevinfluegel]

I can confirm this.

[wkusaa]

can attest to this. it's happening on my opencode cli too

[shanepadgett]

Like literally wont even start working until 10 minutes have past. No indication of throttling or anything. 5.2-Codex is just sitting there doing nothing, then it runs one command, then waits another 10 minutes. I'm actually having issues across multiple providers. OpenAI too. OpenCode seems to just not be processing requests today, but I have a hard time believing its anything to do with OpenCode. Maybe just MS infra?

[mdluo]

The same happening to me. I thought it was just slow during peak hours, but once I switch back to copilot in VS Code, it went back to normal speed. I have oh-my-opencode plugin installed also btw.

[mdluo]

After installing the latest oh-my-opencode (bunx oh-my-opencode@v3.0.0-beta.9 install) and opencode auth logout then /connect again, it seems fixed the issue.

[AMTourky]

It's super slow!! it runs a grep command then keeps waiting for a couple mins to move on!
It feels like there are kind of throttling or rate limiting once using it for long!

[althafdemiandra]

tried in vscode copilot chat and its also superslow
already like 5min+

[shanepadgett]

Yea I was using codex all day in vscode chat (copilot). Fast and responsive, but I just dont like the agent. Back to opencode, soooooo slow. Minutes between a read of a file and then next tool call.

Edit: It seems specific to the 5.2 codex model. Switching to regular 5.2 is much faster.

[rekram1-node]

For the openai models, maybe try now?

For the other models (anthropic/google) will update

[althafdemiandra]

from what i've been experiencing it seems like it is my ISP issues
model from copilot provider working normally and fast as is when im using Cloudflare WARP and immediately got performance degradation right after turned it off