Skip to content

Error model calling #112

Closed
Closed

Description

@ApexGP

Describe the bug

After invoking a session, I start typing prompts and get outputs below:

 × Model call failed: {"message":"The requested model is not
   supported.","code":"model_not_supported","param":"model","type":"invalid_request_error"}

Affected version

Version 0.0.328 · Commit 3755a93

Steps to reproduce the behavior

  • Start copilot cli with command copilot --banner
  • Type hello

Expected behavior

Response received.

Additional context

  • Shell: WindowsPowerShell v5.1
  • Operating System: Windows 11 24h2

Activity

  1. added theissue type on Sep 29, 2025
  2. scarletkc commented on Sep 29, 2025

    @scarletkc

    same issue here

  3. shiquwucan commented on Sep 29, 2025

    @shiquwucan

    Model call failed: {"message":"The requested model is not
    supported.","code":"model_not_supported","param":"model","type":"invalid_request_error"}

  4. despinozap commented on Sep 29, 2025

    @despinozap

    I'm facing a similar issue and realized it started when MCP servers were added.
    After viewing this I tried it disabling the banner due to I had it set to always in config but even without the banner it keeps failing.
    In my case it's solved when MCP servers are removed, but still can't use them 👎

  5. ApexGP commented on Sep 30, 2025

    @ApexGP
    Author

    I'm facing a similar issue and realized it started when MCP servers were added. After viewing this I tried it disabling the banner due to I had it set to always in config but even without the banner it keeps failing. In my case it's solved when MCP servers are removed, but still can't use them 👎

    I've managed to locate the bug— it's actually handled by the model provider Anthropic who now implements an extremely strict restriction on IP address. The only solution is to enable TUN mode inside your proxy utility and switch to a pure host on your PC. And remember to set $env: HTTP_PROXY and $env: HPPTS_PROXY in your shell.

  6. despinozap commented on Sep 30, 2025

    @despinozap

    I'm facing a similar issue and realized it started when MCP servers were added. After viewing this I tried it disabling the banner due to I had it set to always in config but even without the banner it keeps failing. In my case it's solved when MCP servers are removed, but still can't use them 👎

    I've managed to locate the bug— it's actually handled by the model provider Anthropic who now implements an extremely strict restriction on IP address. The only solution is to enable TUN mode inside your proxy utility and switch to a pure host on your PC. And remember to set $env: HTTP_PROXY and $env: HPPTS_PROXY in your shell.

    Nice! It makes sense, since the cli is using sonnet by default. Isn't it?.
    I'll try it as soon as I can.

  7. RyanHecht commented on Oct 1, 2025

    @RyanHecht
    Collaborator

    Hi folks, we improved the /model picker in 0.0.331 to only show you the models your account has access to! Working with our docs team to document the reasons (copilot plan, geographic region, etc) some models may be missing to avoid further confusion.

  8. scarletkc commented on Oct 1, 2025

    @scarletkc

    Hi folks, we improved the /model picker in 0.0.331 to only show you the models your account has access to! Working with our docs team to document the reasons (copilot plan, geographic region, etc) some models may be missing to avoid further confusion.

    Why can my account only use GPT-5 and not Claude 4.5, and it's not the Codex version? I can use Sonnet in VSCode.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions