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
{{ message }}
Repository navigation
[FEATURE]: Support for Claude Opus 4.6 Fast Mode #12923
I have verified this feature I'm about to request hasn't been suggested before.
Describe the enhancement you want to request
First, thank you for maintaining this project. I really appreciate all your work.
I wanted to bring to your attention that both Anthropic and GitHub Copilot have recently announced experimental support for Fast Mode with the Claude Opus 4.6 model. I would love to see this capability available in OpenCode, but it doesn’t appear that such a feature or commit exists yet, and I haven’t seen any related issue requests.
It would be fantastic if this functionality could be supported.
Thank you!
Reacted by Yassine Erraqui, TB, Álvaro Veiga, Lucas Lagrimante Martinho, Fredrik A. Madsen-Malmo, Armando Herrera, Abdellatif Abdelfattah, randrei, Sergter, JSCOP and 7 more
Fast Mode is a specific capability of Opus 4.6. Once Opus 4.6 support is added (especially for GitHub Copilot), the Fast Mode feature would likely be available as part of that implementation.
(Note: I'm writing this from memory without my work environment in front of me, so some details might not be perfectly accurate.)
Yesterday, after briefly doing some LLM-assisted research and analyzing the microsoft/vscode-copilot-chat extension, I discovered that the Copilot domain and /models endpoint used by OpenCode simply do not expose the claude-opus-4.6-fast model at all.
When I requested /models against the domain OpenCode uses (api.githubcopilot.com), the claude-opus-4.6-fast model was not in the list. As a result, all requests fail with the message: "The requested model is not supported."
However, the model selector in VSCode's official Copilot Chat extension clearly includes Opus 4.6 Fast Mode. Digging into the source code based on that, I found that VSCode actually sends requests to a different domain: api.individual.githubcopilot.com. The /models endpoint on this domain does expose the Fast Mode model.
I still need to investigate this further, but I suspect the authentication workflow itself might be different, and it's unclear if GitHub Copilot intended this behavior or not. Just to be sure, I even tried signing out of GitHub Copilot in OpenCode and authenticating again, but the model still didn't appear in the list.
However, the model selector in VSCode's official Copilot Chat extension clearly includes Opus 4.6 Fast Mode. Digging into the source code based on that, I found that VSCode actually sends requests to a different domain: api.individual.githubcopilot.com. The /models endpoint on this domain does expose the Fast Mode model.
@utapyngo Okay, I just don't want to bother them. But come to think of it, GitHub Copilot is one of the official providers for OpenCode, isn't it? So I think I need to reopen this. They might even be able to request GitHub Copilot to support Fast Mode for OpenCode as well.
selecting opus 4.6 from github copilot seems to be using the fast model by default and charging 30x credit, i came across a similar post in reddit. a fast selector would be nice as i donot want 10% of my credits to be gone in an instant.
I think this feature might become urgent. Background: Copilot seems to rate-limit Opus 4.6 quite hard. But the fast version is obviously performing full-speed ( "paying4speed")
Activity
github-actions commentedon Feb 10, 2026
This issue might be a duplicate of existing feature requests. Please check:
Fast Mode is a specific capability of Opus 4.6. Once Opus 4.6 support is added (especially for GitHub Copilot), the Fast Mode feature would likely be available as part of that implementation.
dragonhuntr commentedon Feb 17, 2026
bump on this
utapyngo commentedon Mar 3, 2026
There is an issue on this at models.dev: anomalyco/models.dev#833
Tetrahedrite commentedon Mar 3, 2026
@utapyngo Thanks for letting me know. I didn't realize this was an issue with that repository. I'll go ahead and close this.
utapyngo commentedon Mar 4, 2026
Actually I would keep this open since it seems that changes to opencode are also required. I tried adding this to my
opencode.json:The model appeared in the list but the server keeps responding with "The requested model is not supported".
Tetrahedrite commentedon Mar 5, 2026
@utapyngo
(Note: I'm writing this from memory without my work environment in front of me, so some details might not be perfectly accurate.)
Yesterday, after briefly doing some LLM-assisted research and analyzing the
microsoft/vscode-copilot-chatextension, I discovered that the Copilot domain and/modelsendpoint used by OpenCode simply do not expose theclaude-opus-4.6-fastmodel at all.When I requested
/modelsagainst the domain OpenCode uses (api.githubcopilot.com), theclaude-opus-4.6-fastmodel was not in the list. As a result, all requests fail with the message: "The requested model is not supported."However, the model selector in VSCode's official Copilot Chat extension clearly includes Opus 4.6 Fast Mode. Digging into the source code based on that, I found that VSCode actually sends requests to a different domain:
api.individual.githubcopilot.com. The/modelsendpoint on this domain does expose the Fast Mode model.I still need to investigate this further, but I suspect the authentication workflow itself might be different, and it's unclear if GitHub Copilot intended this behavior or not. Just to be sure, I even tried signing out of GitHub Copilot in OpenCode and authenticating again, but the model still didn't appear in the list.
utapyngo commentedon Mar 5, 2026
So we need to reopen this issue.
Tetrahedrite commentedon Mar 5, 2026
@utapyngo Okay, I just don't want to bother them. But come to think of it, GitHub Copilot is one of the official providers for OpenCode, isn't it? So I think I need to reopen this. They might even be able to request GitHub Copilot to support Fast Mode for OpenCode as well.
Xynonners commentedon Mar 7, 2026
Would love to see this for GPT 5.4 fast mode as well!
gauravyad69 commentedon Mar 15, 2026
selecting opus 4.6 from github copilot seems to be using the fast model by default and charging 30x credit, i came across a similar post in reddit. a fast selector would be nice as i donot want 10% of my credits to be gone in an instant.
V4nus commentedon Mar 23, 2026
please
tisDDM commentedon Apr 3, 2026
I think this feature might become urgent. Background: Copilot seems to rate-limit Opus 4.6 quite hard. But the fast version is obviously performing full-speed ( "paying4speed")
Would be great to see an improvement here.