Feature Request: Add support for Qwen 3.5 series models (including Qwen3.5-0.8B/2B/4B/9B) #479
Description
Is your feature request related to a problem? Please describe.
Currently, Foundry Local provides a great experience with the Qwen 2.5 family. However, Alibaba has recently released the Qwen 3.5 series, which brings significant improvements in reasoning, coding, and multilingual capabilities. As a developer using Intel Arc GPU (specifically the B580 Battlemage), I'm eager to leverage these new models natively via the OpenVINO backend in Foundry Local for better performance.
Describe the solution you'd like
I would like to see the following Qwen 3.5 models added to the Foundry Local model catalog, optimized for OpenVINO (CPU/GPU/NPU):
- Qwen3.5-0.8B-Instruct: Ideal for ultra-low latency edge tasks.
- Qwen3.5-2B-Instruct: A perfect balance for mobile and lightweight local dev.
- Qwen3.5-4B-Instruct: Stronger logic while maintaining high speed.
- Qwen3.5-9B-Instruct: The new sweet spot for high-quality local reasoning.
- Qwen3.5-35B-A3B (MoE): High-performance sparse model for more complex tasks.
Describe alternatives you've considered
I have tried running these via Ollama or manual OpenVINO GenAI implementation, but Foundry Local's seamless integration with Windows and its API-first approach makes it the preferred tool for my C#/.NET development workflow.
Additional context
The Qwen 3.5 series models are already gaining traction in the community (Hugging Face / Unsloth), and having them available as "out-of-the-box" options in Foundry Local would significantly benefit developers on the Windows AI PC ecosystem, especially those using the latest Intel Battlemage hardware.
