DeepSeek's New Visual Model deepseek-v4-flash-vision-exp Enters Official Harness

The new visual model deepseek-v4-flash-vision-exp from DeepSeek has been added to the official Harness default model directory, supporting image input. Community users have tested this model via the API, allowing them to directly upload images for inquiries. The code PR title for this model is "publish the vision model." Currently, DeepSeek has not officially announced this, and the official API documentation has not been updated with related information. DeepSeek has long had image recognition capabilities, with both the website and app already featuring image recognition mode. Previously, the research by the DeepSeek visual team was based on V4-Flash, integrating points and bounding boxes directly into the inference chain to achieve visual CoT inference. Two days ago, DeepSeek Harness RC.8 added native image input functionality to the official adapter, although deepseek-v4-flash-vision-exp was written at that time, it was not included in the default model directory due to waiting for the model endpoint to be ready. This restriction has now been lifted, and Harness v0.1.1-rc.1 lists deepseek-v4-flash-vision-exp as the default visual model, marking the imminent release of the new visual model.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.