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
Same issue with GH Copilot and Sonnet 4.5, Haiku 4.5.
Switched to Gemini and it worked.
Reacted by damiankaplonspa, Mohamed Yassine Landolsi , Yinzhou Jiancheng, Aleksandar Janković, TmacAaron, Begüm Irmak Özcan, Sighthesia, fynn and Hugo Labraaten
Seeing the issue with Codex 5.2 with OpenAI as the provider as well - it seems like the foreground agent is unable to write files, but delegated sub-agents don't seem to have the issue for me.
Session is yet another task manager for my taste so not important, here you go -> session.json
This could be a classic case of no loading bar / mirror in the elevator. Have you considered adding a token count next to "preparing XXX" to see the model writing down tokens? That may be the only issue that exists
Session is yet another task manager for my taste so not important, here you go -> session.json
This could be a classic case of no loading bar / mirror in the elevator. Have you considered adding a token count next to "preparing XXX" to see the model writing down tokens? That may be the only issue that exists
Just tested, it is indeed the case! Be patient enough and it works: session.json
PR 16024 saw exactly that and fixed it, but it got lost in all the PRs. @rekram1-node , can you reopen it ?
This issue #11112 about extreme slowness when using the Opus model was created back in January, and it’s already July now—half a year has passed without any proper fix or meaningful progress update. I’m really confused about the maintenance status of this repo. Could maintainers explain why this critical performance bug affecting all Opus users has been left unresolved for so long? Many of us rely on Opus for complex coding work, and the constant long loading times and hanging executions seriously ruin daily development efficiency. We’d appreciate a clear timeline or roadmap for this fix.
For me, at the end it was giving an error and no output. I was charged, but no work was done. So I moved to pi and I'm happy. Will try again opencode when they fix that issue.
@YurySolovyov
I create a plugin that lets you view the current status (including editing status & tokens) in the sidebar. Although the user experience isn't great, it's better than nothing.
@Sun-ZhenXing curious if there is a way to show max timeout in bash tool, so it looks like: bash · 12.5s / 30s or something, because sometimes models request larger timeouts. Otherwisw looks very cool!
Closing as a duplicate of #18108. Getting stuck on "Preparing write..." usually means the tool call was cut off at the model's output token limit, and OpenCode doesn't report that clearly or recover from it. Please follow and 👍 there.
Activity
github-actions commentedon Jan 29, 2026
This issue might be a duplicate of existing issues. Please check:
Feel free to ignore if none of these address your specific case.
marcinpz commentedon Jan 29, 2026
Same here:
vidyanand commentedon Jan 29, 2026
I am facing the same issues as well. I am using 1.1.40 version and am using claude-sonner through github copilot.
andriy-ah commentedon Jan 29, 2026
Same issue with GH Copilot and Sonnet 4.5, Haiku 4.5.
Switched to Gemini and it worked.
Ylandolsi commentedon Jan 29, 2026
same thing here
vaguilarrebolledo commentedon Jan 29, 2026
same issue with claude-opus-4.5
rekram1-node commentedon Jan 29, 2026
Going to investigate
allen-n commentedon Jan 29, 2026
Seeing this as well, some debug info:
model provider: github copilot, opus 4.5
vaguilarrebolledo commentedon Jan 29, 2026
i connected with github copilot provider
spoons-and-mirrors commentedon Jan 29, 2026
are you guys using any plugin ?
allen-n commentedon Jan 29, 2026
@spoons-and-mirrors No, but it does seem like we are all using copilot
allen-n commentedon Jan 29, 2026
Seeing the issue with Codex 5.2 with OpenAI as the provider as well - it seems like the foreground agent is unable to write files, but delegated sub-agents don't seem to have the issue for me.
rekram1-node commentedon Jan 29, 2026
Can anyone export a session to me? Or steps to replicate?
opencode export > session.json^^ run this command in ur terminal and then send file
66 remaining items
victorhramos-dev commentedon Jun 3, 2026
same here...
utorque commentedon Jun 23, 2026
Still happening, 1.17.9
Session is yet another task manager for my taste so not important, here you go ->
session.json
This could be a classic case of no loading bar / mirror in the elevator. Have you considered adding a token count next to "preparing XXX" to see the model writing down tokens? That may be the only issue that exists
utorque commentedon Jun 23, 2026
Just tested, it is indeed the case! Be patient enough and it works:
session.json
PR 16024 saw exactly that and fixed it, but it got lost in all the PRs. @rekram1-node , can you reopen it ?
wenbochang888 commentedon Jul 8, 2026
This issue #11112 about extreme slowness when using the Opus model was created back in January, and it’s already July now—half a year has passed without any proper fix or meaningful progress update. I’m really confused about the maintenance status of this repo. Could maintainers explain why this critical performance bug affecting all Opus users has been left unresolved for so long? Many of us rely on Opus for complex coding work, and the constant long loading times and hanging executions seriously ruin daily development efficiency. We’d appreciate a clear timeline or roadmap for this fix.
YurySolovyov commentedon Jul 12, 2026
Perhaps patch tool decorations might show a counter of tokens or lines, so people know that it is not stuck? e.g:
lorenc-tomasz commentedon Jul 13, 2026
For me, at the end it was giving an error and no output. I was charged, but no work was done. So I moved to
piand I'm happy. Will try again opencode when they fix that issue.mazuralexandru commentedon Jul 21, 2026
The issue still persists. Please add a progress indicator so we know it hasn't just frozen
coder-czw commentedon Aug 12, 2026
still persists in opencode@1.18.16 , any update for this issue?
Sun-ZhenXing commentedon Aug 14, 2026
@YurySolovyov
I create a plugin that lets you view the current status (including editing status & tokens) in the sidebar. Although the user experience isn't great, it's better than nothing.
Try:
~/.config/opencode/opencode.json:{ "plugin": ["@alexsun-top/opencode-turbo@0.1.0"] }and:
~/.config/opencode/tui.json{ "plugin": ["@alexsun-top/opencode-turbo@0.1.0"] }YurySolovyov commentedon Aug 14, 2026
@Sun-ZhenXing curious if there is a way to show max timeout in bash tool, so it looks like:
bash · 12.5s / 30sor something, because sometimes models request larger timeouts. Otherwisw looks very cool!Sun-ZhenXing commentedon Aug 14, 2026
@YurySolovyov This feature is already supported. Try v0.1.1.
kennylbj commentedon Sep 8, 2026
Still a issue in opencode@1.18.29, any progress on fixing this?
DSK1113205 commentedon Sep 29, 2026
The issue occurs when the context length exceeds its limit while editing files. Attention Plss - opencode v2.0.19
neriousy commentedon Oct 1, 2026
Closing as a duplicate of #18108. Getting stuck on "Preparing write..." usually means the tool call was cut off at the model's output token limit, and OpenCode doesn't report that clearly or recover from it. Please follow and 👍 there.