Add Qwen3 NPU optimized model with Tools Support

[zytoh0]

Hello team,

I'd like to propose the addition of the Qwen3 series of NPU-optimized models into the AI model catalog utilized by the VS Code AI Toolkit, specifically focusing on models optimized for agentic and tool-augmented workflows.

Why This Matters:

The Qwen3 models represent a significant leap forward in reasoning, multilingual support, coding, mathematics, and agent integration capabilities. Optimizing these models for NPU (Neural Processing Unit) deployment would enable developers to leverage their full potential in edge environments and limited-resource devices, greatly enhancing performance and user experience.

Particularly, prioritizing the largest Qwen3 model that can fit into NPU memory ensures maximizing capability without sacrificing operational efficiency.

Feature Request Details:

✅ Add Qwen3 series models to the supported model list, including multiple sizes (e.g., 4B, 8B, 14B, 30B, 32B). ✅ Focus on NPU-optimized variants, prioritizing the largest possible model per memory constraints. ✅ Ensure full Tools Support (function calling, external tool usage, agent-based interaction). ✅ Provide clear model card/documentation outlining:

• Quantization methods used (e.g., Q4_K_M, Q8, etc.)
• Constraints (e.g., minimum hardware requirements, memory footprint)
• Examples to run thinking mode vs. non-thinking mode for optimal task execution

References:

• 🦙 Ollama Qwen3 Library
• 🤗 Hugging Face Qwen3 Collection
• 📂 Qwen3 Official GitHub Repository
• 📰 Qwen3 Official Blog Announcement

[microsoft-github-policy-service[bot]]

Thank you for contacting us! Any issue or feedback from you is quite important to us. We will do our best to fully respond to your issue as soon as possible. Sometimes additional investigations may be needed, we will usually get back to you within 2 days by adding comments to this issue. Please stay tuned.

[hi-brenda]

Thank you for your feedback. We've added this to our backlog for future consideration. While we can’t commit to a timeline right now, your input helps us prioritize improvements.