GPT 5.6 Sol is the best "vision" model OpenAI ever released

OpenAI's GPT-5.6 Sol model represents a major advancement in vision capabilities, particularly in object detection where it achieved 46.2 mAP@50 compared to GPT-5.5's 13.8 mAP@50. The model shows strong performance across detection, counting, OCR, and data extraction tasks, though it exhibits some instability on larger images (2,000+ pixels), occasionally placing bounding boxes in incorrect locations. Terra and Luna models in the lineup also show meaningful improvements over their predecessor, making vision a practical capability for these models.