Which DeepSeek V4 models accept images? OpenRouter breaks down the family

DeepSeek V4 哪些型号支持图像输入？OpenRouter 逐款梳理与调用指南

DeepSeek V4 is a model family, not a single model. OpenRouter's guide confirms only V4.1 Flash and V4 Flash Vision Exp accept image input; all others (V4 Pro 0813, V4 Flash 0731, etc.) are text-only. V4.1 Flash is the recommended choice with native vision support at $0.15/$0.60 per million tokens. V4 Flash Vision Exp is the pricier experimental option. The post also covers two integration methods: direct image input or using a separate vision model as a front-end.