Qwen Models for GitHub Copilot Chat

Use Qwen Token Plan / DashScope models directly in Copilot Chat as a BYOK provider extension.

Features

• Adds a Qwen provider in Copilot model picker
• Supports Token Plan text models (Qwen/DeepSeek/GLM) with per-model context windows (e.g. 983K for qwen3.8/3.7 flagships, 1M for GLM 5.x) and auto-discovery of model IDs from /models that inherits the built-in specs
• Thinking effort picker (none/high/max) for thinking-capable models; requests send Qwen's native enable_thinking switch
• Bilingual UI (English + 简体中文) for commands, settings, and usage panel
• Token Plan usage status bar + detail panel: 5-hour / weekly / monthly quota with reset times and plan name
• Custom endpoint and custom model support

Setup

1. Open Command Palette and run Qwen: Set API Key
2. Select mode in settings:
   • qwen-copilot.apiMode = coding-plan for Token Plan
   • qwen-copilot.apiMode = standard for DashScope pay-as-you-go
3. Choose qwen-copilot.region:
   • Token Plan default: token-plan.maas.qianwenaiapi.com/compatible-mode/v1
   • Standard pay-as-you-go: dashscope-intl.aliyuncs.com/compatible-mode/v1 or dashscope.aliyuncs.com/compatible-mode/v1

Optional:

• Set qwen-copilot.baseUrl for a custom compatible gateway
• Set qwen-copilot.tokenPlanStatsEndpoint if your usage endpoint differs
• Disable qwen-copilot.enableModelDiscovery if you prefer static model list only

Usage visibility

The status bar queries Token Plan usage through the same gateways the official Qwen CLI uses:

1. cs-data gateway (cs-data.qianwenai.com) — monthly usage percentage, Bearer auth. If the gateway rejects the API key, the client automatically falls back to:
2. Alibaba console RPC (queryCodingPlanInstanceInfoV2 on Model Studio / Bailian consoles) — 5-hour, weekly, and monthly quota with reset times, authenticated with your API key and region-aware commodity codes.

If both are unavailable, the panel shows No data instead of a hard failure. Click the status bar item to open the detail panel. A custom override endpoint is still supported via qwen-copilot.tokenPlanStatsEndpoint (both response shapes are detected automatically).

Notes

• This extension stores API keys in VS Code SecretStorage.
• For latest endpoint and plan policy changes, check official docs:
  • Qwen AI Platform Token Plan: https://platform.qianwenai.com/pricing/token-plan
  • DashScope OpenAI-compatible endpoint guide: https://www.alibabacloud.com/help/en/model-studio/compatibility-of-openai-with-dashscope