This new LFM 2.5 VL model at 450 B is amazing! I have it in Perspective Intelligence after working with Codex to support in with MLX, and it is amazing! less than 2 second analysis times.