THE MODELS DESK

Models.

The capabilities, trade-offs and people behind China’s AI models.

70 storiesPage 10 of 12

WeChat tests AI image editing and information extraction

Some WeChat users are seeing tools for image modification, text extraction, translation and summarization in the image-sending flow.

Brief2 min readReported Sep 14

A subset of DeepSeek users can have responses read aloud and choose between four voice profiles.

Brief2 min readReported Sep 14

DeepSeek's Harness 0.1.5 release candidates add V4.1 Flash support, file previews and more control over continuing agent work.

Brief2 min readReported Sep 10

V4.1 Flash adds a lower-cache multimodal architecture, while updated API documentation keeps V4 Pro available separately.

News3 min readReported Sep 10

Bloomberg reporting relayed by TechNode describes a ByteDance project aimed at generating responsive virtual environments.

Brief2 min readReported Sep 8

A limited WeChat experiment lets assistants handle preliminary coordination before people make the important decisions.

Brief2 min readReported Sep 8