Woofun AI reports that DeepSeek has integrated the deepseek-v4-flash-vision-exp model into the default directory of its official Harness tool. This update explicitly marks the model as supporting image input, following a code pull request titled 'publish the vision model'.

Although no official announcement or API documentation update has been released, community users have already accessed the model via APIs to upload images. The integration follows the lifting of previous restrictions in Harness RC.8, indicating that the new visual model is prepared for broader release.