ToolsAugust 21, 2026via Vercel Blog
DeepSeek V4 Flash with vision now available on AI Gateway
Why it matters
A capable vision-language model variant is now accessible through a major AI platform integration, expanding multimodal options for practitioners building with agents and APIs. The 1M context window and tool use keep the Flash efficiency while adding image understanding.
Key signals
- DeepSeek V4 Flash vision available on Vercel AI Gateway
- Experimental release (marked -exp); context window 1M tokens, output 384K tokens
- Supports JPEG, PNG, GIF, WebP image formats
- Tool use, reasoning, and implicit caching functional
- Integrates with coding agents: Claude Code, Codex, OpenCode, Cursor, Pi
- Date: August 21, 2026
- DeepSeek V4 Flash with vision now live on Vercel AI Gateway
- Supports JPEG, PNG, GIF, WebP images in same request as text
- Context window: 1M tokens input, 384K output (unchanged from base Flash)
- Tool use, reasoning, implicit caching supported
- Integration ready for Claude Code, Codex, OpenCode, Cursor, Pi and other agents via vercel ai-gateway coding-agents setup
- Marked as -exp (experimental); model ID subject to change
- Format detection from file bytes, not extension or mediaType declaration
The hook
DeepSeek V4 Flash adds vision to AI Gateway—multimodal reasoning now available for coding agents and workflows.
DeepSeek V4 Flash with vision is now available on AI Gateway.
This model is an experimental version that accepts images alongside text. You can ask it to describe a picture, read text out of a screenshot, or work through a chart, in the same request as your prompt. It keeps what the Flash line alrea…