DeepSeek's whale opens its eyes in tease of new vision model

DeepSeek multimodal team researcher Xiaokang Chen (@PKUCXK on Telegram) posted a cryptic message on X: "Now, we see you. 👀" The post includes a comparison of two DeepSeek whale mascots — on the left, the whale's eyes are covered; on the right, they're wide open.

Given Chen's bio — "Researcher in Multimodal Team @deepseek_ai" — and the "seeing" metaphor, it strongly suggests DeepSeek is about to launch a brand-new multimodal (vision) large model.

DeepSeek V4, released on April 24, is a text-only model that doesn't support image input. A multimodal model would be a separate product line. DeepSeek has made no official announcement yet. If it materializes, it would join a growing field of multimodal models, including Alibaba's Qwen-Image-2.0-Pro, which generates and edits images with multilingual text.