🐛 Bug Report: Vertex AI Qwen Model ID Fails to Parse in Goose CLI

When attempting to start a Goose session using the qwen/qwen3-coder-480b-a35b-instruct-maas model via Google Cloud Vertex AI, the CLI fails with an error related to parsing the model ID. Authentication with GCP is working correctly, and other models like gemini-2.0-flash-001 work without issue

This seems to be a problem either with how the model ID is parsed or an unsupported model path formatting.

💥 Actual Behavior
Steps to reproduce the behavior:

1. Ensure GCP authentication is correctly configured and working.
2. Update the Goose config file at: ~/.config/goose/config.yaml
   With the following content:

GCP_LOCATION: us-central1
GCP_PROJECT_ID: your-project-id
GOOSE_MODEL: qwen3-coder-480b-a35b-instruct-maas # or qwen/qwen3-coder-480b-a35b-instruct-maas
GOOSE_MODE: auto
GOOSE_CLI_MIN_PRIORITY: -1
GOOSE_PROVIDER: gcp_vertex_ai

3. Start Goose CLI: goose s
4. See error

✅ Expected behavior
The session should start successfully with the selected Qwen model, similar to how it works with gemini-2.0-flash-001, and respond to prompts.

🧰 Environment:

• OS & Arch: MacOS Sequoia 15.6 M4
• Interface: CLI
• Version: 1.4.1
• Extensions enabled: context7, developer, memory, perplexityai, taskmaster,
• Provider & Model: GCP Vertex: qwen/qwen3-coder-480b-a35b-instruct-maas

🔗Model Reference
Here is the model page on Vertex AI Model Garden for reference:
Qwen3 Coder Model on GCP Vertex AI

🖼️ Screenshots