Video Generation, Vision / UI-UX / Screenshot-to-Code Models, Coding Models,

[EswarVarma91]

Source link

https://huggingface.co/Wan-AI, https://huggingface.co/tencent/HunyuanVideo, https://huggingface.co/Lightricks, https://huggingface.co/THUDM/CogVideoX-5B, https://huggingface.co/hpcai-tech, https://huggingface.co/Qwen, https://huggingface.co/deepseek-ai/DeepSeek-V3, https://huggingface.co/deepseek-ai/DeepSeek-R1, https://huggingface.co/mistralai, https://huggingface.co/meta-llama, https://huggingface.co/OpenGVLab, https://huggingface.co/allenai

Target platform

macOS

Use case

For video generation, support for Wan 2.1, HunyuanVideo, LTX Video, and CogVideoX would be valuable. For coding and AI agents, support for DeepSeek V3, Qwen3 Coder, DeepSeek R1, and OpenHands would enable powerful local software development workflows. For UI/UX design and screenshot-to-code generation, support for Qwen2.5-VL, Qwen3-VL, InternVL, and related multimodal models would allow developers and designers to build applications directly on Apple hardware. Native optimization for Apple Silicon (M-series chips) would help users run these models privately, securely, and without ongoing costs, making Apple devices a leading platform for local AI development.

Preferred precision / compression

4-bit

Additional context

No response

[stikves]

Thank you for the suggestions.

Yes, video generation would be a useful feature.
Similar with VLMs.

While we cannot provide a timeline, those two features are in consideration for future potential expansion.

We already support Qwen-3 coder in our current models:
https://github.com/apple/coreai-models/tree/main/models/qwen3_moe