DeepSeek V4 Flash Vision Model Likely Imminent

According to X user MaxForAI, signs of a DeepSeek vision model have emerged. At 3 PM today, the name "deepseek-v4-flash-vision-exp" appeared in DeepSeek Harness-related code, and the latest Harness version has begun adding native image request support. However, the model will not be added to the official model catalog until its endpoint is ready. At present, DeepSeek appears to be testing the vision version of V4 Flash, with the public API not yet officially available. Some users have already tried requesting the DeepSeek API directly with this model ID, but it still returns errors, and the model is not listed in official documentation. The adaptation paths on both the client and Harness sides are largely in place, and DeepSeek V4 Flash is expected to gain vision capabilities soon.