DeepSeek releases V4-Flash-Vision-Exp multimodal model on API

DeepSeek launched the experimental V4-Flash-Vision-Exp multimodal model on its API platform, matching V4-Flash text performance while adding image understanding that brings multimodal agent benchmarks close to Anthropic Opus-4.8 levels. Images are billed at V4-Flash rates (up to 384 tokens each) and the model supports mixed text-image input via base64, URLs or the new free Files API. The release extends a low-cost Chinese open-weight lineage into practical vision-agent workflows.