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

	GPT 5.6 Sol is the best "vision" model OpenAI ever released(blog.roboflow.com)
	359 points by plurby 4 days ago \| 166 comments
	tl;dr: OpenAI's new GPT-5.6 lineup (Sol, Terra, Luna) shows major gains in vision tasks, with Sol jumping from 13.8 to 46.2 mAP@50 on object detection and improving counting across the board, though OCR performance stayed flat or slightly regressed. Sol becomes unstable on images near 2000x2000px and costs ~2.5 cents/image with ~10s latency, making Gemini 3.5 Flash still the better price/performance choice for high-volume detection and counting workloads.
	HN Discussion: ↑Article understates that Gemini 3.5 Flash clearly beats Sol on nearly all benchmarks at lower cost ~Sol performs well in practical vision tasks like UI analysis based on personal experience ~Using LLM vision models for simple tasks like pill counting is wasteful when traditional CV solves it ↓Benchmark methodology has flaws like EXIF orientation errors and incorrect expected answers ↓Comparison is incomplete without including other Gemini versions like 3 or 3.7