[Bug] Models list shows only "Auto (Copilot)", selecting it throws "Resolved model not found" error

[chungyinleo]

I am experiencing an issue where CopilotChat.nvim cannot find or use any of the standard models (Gemini, Claude, GPT), even though my core Copilot configuration is fully online and inline suggestions are working perfectly.

Current Behavior:

1. Running :CopilotChatModels displays only one option: 1. Auto (Copilot) [copilot].
2. Selecting that option and trying to use the chat causes the plugin to crash with a Resolved model not found error.

The core copilot.lua extension is verified online, and code completions work normally, so this does not appear to be an issue with my GitHub subscription or basic Copilot connectivity. Re-authenticating did not solve the issue.

Steps to Reproduce:

1. Ensure copilot.lua is authenticated and actively providing inline suggestions.
2. Run :CopilotChatModels to see only the single fallback option.
3. Open CopilotChat and send a message using that model.

Error Logs:

[ERROR] Resolved model not found: gpt-5.4-mini                                                                                                   
stack traceback:                                                                                                                                 
^I...hare/nvim/lazy/CopilotChat.nvim/lua/CopilotChat/init.lua:201: in function <...hare/nvim/lazy/CopilotChat.nvim/lua/CopilotChat/init.lua:198> 
^I[C]: in function 'error'                                                                                                                       
^I...re/nvim/lazy/CopilotChat.nvim/lua/CopilotChat/client.lua:328: in function 'ask'                                                             
^I...hare/nvim/lazy/CopilotChat.nvim/lua/CopilotChat/init.lua:539: in function <...hare/nvim/lazy/CopilotChat.nvim/lua/CopilotChat/init.lua:468> 
^I[C]: in function 'xpcall'                                                                                                                      
^I...hare/nvim/lazy/CopilotChat.nvim/lua/CopilotChat/init.lua:205: in function <...hare/nvim/lazy/CopilotChat.nvim/lua/CopilotChat/init.lua:197> 

This error is from my Linux desktop, but the same happen on my Mac too at essentially the same time, without updating any neovim plugins. My github overview indicates I am not exceeding the quota.

Environment / System Info:

• OS: Linux (NixOS)
• Neovim Version: 0.12.3 (I am using Lazyvim)
• CopilotChat.nvim Branch: main

[pullfrog]

  ^{｜ Build this ➔ ｜ Make a plan ➔}

[IqbalHossainEmon]

I think it's an issue with GitHub Copilot updating their model policy for the student pack and free users, I'm also facing. this on a Free/Student plan account too — same failure mode, with gpt-5.4-mini as the resolved model in my case as well.

Environment:
Neovim 0.12.3
CopilotChat.nvim installed via vim.pack, branch main, updated today
copilot.lua for inline suggestions (working fine, unaffected)
Account: GitHub Copilot Free/Student plan (auto-only model selection, per the June 24, 2026 policy change)

What I'm seeing:

:CopilotChatModels shows only:

• Auto (Copilot) [copilot]

With model = 'auto' set explicitly, sending any message fails with Resolved model not found: .

Debug log (tokens redacted):

[CopilotChat] [DEBUG] POST request: https://api.individual.githubcopilot.com/models/session {
  body = {
    auto_mode = {
      model_hints = { "auto" }
    }
  },
  headers = {
    Authorization = "<redacted>",
    ["Copilot-Integration-Id"] = "vscode-chat",
    ["Editor-Version"] = "Neovim/0.12.3",
    ["x-github-api-version"] = "2025-10-01"
  }
}
[CopilotChat] [DEBUG] POST response: https://api.individual.githubcopilot.com/models/session {
  body = '{"available_models":["gpt-5.4-mini","claude-haiku-4.5","gpt-5-mini","gpt-5.3-codex"],"selected_model":"gpt-5.4-mini","session_token":"<redacted>","expires_at":...}',
  status = 200
}
[CopilotChat] [ERROR] Resolved model not found: gpt-5.4-mini
stack traceback:
        .../CopilotChat.nvim/lua/CopilotChat/init.lua:201: in function <.../init.lua:198>
        [C]: in function 'error'
        .../CopilotChat.nvim/lua/CopilotChat/client.lua:328: in function 'ask'
        .../CopilotChat.nvim/lua/CopilotChat/init.lua:539: in function <.../init.lua:468>

My guess:
The auto_mode session handshake itself works correctly — the POST to /models/session returns a 200 with a valid selected_model (gpt-5.4-mini) drawn from a real available_models list. The problem looks like it's at client.lua:328: the resolved selected_model is being checked against the original static /models fetch (the one that populates :CopilotChatModels), which for Free/Student accounts now only contains the single Auto placeholder entry. Since gpt-5.4-mini isn't in that narrower list, the lookup fails even though the session response already confirmed it's a valid, routable model for this account.

It seems like the fix would be to trust selected_model from the /models/session response directly when running in auto mode, rather than re-validating it against the restricted top-level /models list. Happy to test a patch if one gets put up.

[Belal-Abo-Ata]

Any fix for this issue yet?

[druxorey]

I've been experiencing the same issue, hopefully this can be fixed soon.