DeepSeek V4 Flash Vision Experimental now available on AI Gateway
By Steven Van ·
DeepSeek V4 Flash with vision is now available on AI Gateway. This model is an experimental version that accepts images alongside text.
DeepSeek V4 Flash Vision Experimental is now available on Vercel's AI Gateway, adding image input to the existing DeepSeek V4 Flash model. A single request can mix a prompt with a picture, a screenshot, or a chart, and the model will describe it, read text out of it, or work through it alongside the text. Tool use, reasoning, and caching work as they did before.
The model accepts JPEG, PNG, GIF, or WebP files, and reads the format from the file's own bytes rather than its filename or declared media type. It's available as deepseek/deepseek-v4-flash-vision-exp, and can also be selected inside coding agents such as Claude Code, Codex, OpenCode, Cursor, and Pi after running vercel ai-gateway coding-agents setup. The -exp in the model ID marks it as experimental, so Vercel recommends keeping a fallback model configured on any production path. It can be tried in the model playground, and like other AI Gateway models, it's billed at provider pricing with no platform markup, including for BYOK requests.
