 DeepSeek shipped their first vision model: V4 Flash Vision Exp. It's already available in Yao Agents. ## What This Means Until now, if you were using DeepSeek models and your agents needed to analyze a screenshot, read a chart, or review a UI mockup, you had to wire up a separate vision-capable model from another provider. That's no longer the case — DeepSeek now handles text and images in the same model family. ## The New Vision Models | Model | What it does | |-------|-------------| | DeepSeek V4 Flash Vision Exp | Vision + tool calls, optimized for speed | | DeepSeek V4 Flash Vision Exp Thinking High | Vision + extended reasoning for complex visual analysis | Both share the same 1M token context window as V4 Flash. ## Setup Settings → Model Configuration → Add Provider. Choose: - **DeepSeek (OpenAI)** — official API, `api.deepseek.com` - **OpenCode Go** — gateway, `opencode.ai/zen/go` ## Also New | Model | Use case | |-------|----------| | DeepSeek V4 Pro Thinking High | Heavy reasoning, complex multi-step tasks | | DeepSeek V4 Flash | Fast and cheap, good for everyday work | ## Version Requirements - Yao Engine Desktop **1.4.3+** - Yao **1.0.0-rc17+** Grab the update from the [download page](/download) or check for updates inside Yao Engine Desktop.
DeepSeek Can See Now
DeepSeek V4 Flash Vision Exp is live in Yao Agents. Pick it from DeepSeek or OpenCode Go when adding a model provider.