[BUG] VSCode "There's an issue with the selected model" With OpenRouter

[carrotcreamsoup]

Preflight Checklist

• I have searched existing issues and this hasn't been reported yet
• This is a single bug report (please file separate reports for different bugs)
• I am using the latest version of Claude Code

What's Wrong?

When using the the VSCode Claude extension with OpenRouter keys (as described here), I always get this error message, regardless of which model I use:

There's an issue with the selected model (claude-opus-4-6[1m]). It may not exist or you may not have access to it. Run --model to pick a different model.

There's another issue that I thought was related to this, but downgrading to VSCode extension v2.1.77 did not resolve the issues.

When I use claude directly in the terminal, I still get the same issue. But by using /logout, when I restart claude, the issue is gone.

What Should Happen?

VScode Claude extension should work.

Error Messages/Logs

There's an issue with the selected model (claude-opus-4-6[1m]). It may not exist or you may not have access to it. Run --model to pick a different model.

Steps to Reproduce

1. Install Claude Code for VS Code in VS Code and setup an OpenRouter key as described here: https://openrouter.ai/docs/guides/coding-agents/claude-code-integration

Claude Model

Not sure / Multiple models

Is this a regression?

Yes, this worked in a previous version

Last Working Version

No response

Claude Code Version

VSCode extension v2.1.92

Platform

Other

Operating System

Ubuntu/Debian Linux

Terminal/Shell

VS Code integrated terminal

Additional Information

No response

[github-actions]

Found 3 possible duplicate issues:

1. [BUG] Desktop app sends corrupted model name in API URL — ANSI escape codes embedded in endpoint path (404) #44352
2. [BUG] VSCode extension fails on first query after restart with "model may not exist" error #43301
3. [VSCode Extension] Unsupported content type: redacted_thinking when using OpenRouter #37465

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

[niranjandasMM]

Hi @carrotcreamsoup , can I know if how you fixed it, and using the custom Model or openroueter model ?

[carrotcreamsoup]

@niranjandasMM It was with openrouter models. I don't know why, but after a couple days and after updating claude and the Claude VS Code extension, it started working again. Not too sure what happened tbf!

[FelixIsaac]

Seeing same surface error ("There's an issue with the selected model…") when routing Claude Code to a local Ollama Anthropic-compatible API.

Root cause on my side: request was missing Authorization header even w/ ANTHROPIC_AUTH_TOKEN set (debug: has Authorization header: false). Backend then returns 404 model ... not found, CC surfaces the generic "selected model" error.

Workaround: force auth header via ANTHROPIC_CUSTOM_HEADERS:

claude --settings '{"env":{"ANTHROPIC_BASE_URL":"http://localhost:11434","ANTHROPIC_CUSTOM_HEADERS":"Authorization: Bearer ollama"}}' --model gemma4:latest

After this, debug shows has Authorization header: true and requests succeed.

Related locked issue w/ same auth-header symptom: #39013.

[FuturMix]

The root cause @FelixIsaac identified is spot-on — Claude Code sometimes drops the Authorization header when routing through a custom ANTHROPIC_BASE_URL, which causes the upstream proxy/gateway to return a 404 model not found (or similar), and CC surfaces the generic "selected model" error.

A few things that helped me debug similar issues:

1. Verify your proxy actually receives the auth header

On the proxy side, log incoming request headers. If Authorization: Bearer <key> is missing, CC is stripping it before forwarding.

2. The ANTHROPIC_CUSTOM_HEADERS workaround

As Felix showed, you can force the header:

export ANTHROPIC_CUSTOM_HEADERS="Authorization: Bearer YOUR_KEY"

This works for both local proxies (Ollama, vLLM) and remote gateways (OpenRouter, LiteLLM, etc.).

3. Check model name format

When using a gateway/proxy, the model name CC sends must match what the proxy expects. Some proxies expect the full anthropic/claude-sonnet-4 path format (OpenRouter style), others accept claude-sonnet-4 directly. Mismatch → 404 → "model issue" error.

4. Multi-model gateway approach

If you are routing through a gateway like LiteLLM or a managed relay, make sure:

• The gateway's Anthropic-compatible endpoint accepts the model name format CC sends
• The gateway forwards x-api-key or Authorization correctly to the upstream provider
• The gateway returns standard Anthropic response format (not OpenAI format)

The underlying issue is that CC's error handling conflates "model not found" with "auth failure" with "network error" — all surface as the same generic message. Better error propagation would help a lot here.

[alvissone]

this bug has happened again, i rollback VS Code extension version to v2.1.172, it's ok~