LM Studio introduces unified multi-modal MLX engine architecture
Action Required
Users of LM Studio's MLX engine will benefit from improved performance and new multi-modal capabilities, particularly when using vision-enabled models like Gemma 3.
AI Impact Summary
LM Studio has introduced a unified multi-modal engine architecture leveraging `mlx-lm` and `mlx-vlm` to improve performance and user experience when using multi-modal MLX VLMs like Google's Gemma 3. This new architecture seamlessly integrates vision model components as "add-ons" to the core `mlx-lm` text model, enabling features like prompt caching for multi-modal chats and significantly faster follow-up responses. This represents a capability upgrade, enhancing the functionality of LM Studio's MLX engine and expanding its support for vision-enabled LLMs.
Affected Systems
- Date
- Date not specified
- Change type
- capability
- Severity
- high