MLX has released several significant updates and new products. Prince Canuma announced a 23.24% overall speedup for Qwen2-VL on MLX, with specific improvements in prompt speed (17.97%) and generation speed (26.17%). The latest version of mlx-vlm, v0.0.14, includes a refactored KVCache and support for Qwen2-VL. ChatMLX 1.0.0, powered by Apple MLX, has been launched with a new user interface. Additionally, MLX Chat, a chat interface for on-device Language Model use on Apple Silicon built on FastMLX, has been introduced. Future updates will include Trainer (LoRA) and batch processing.