DeepSeek V4 Flash Vision Exp

DeepSeek V4 Flash Vision Exp is an experimental API-only multimodal model released on August 21, 2026. See what makes this AI model special!

Multimodal ModelVision APICoding AgentLong Context

DeepSeek API

Off-peak: $0.007/1M cache-hit input, $0.22/1M cache-miss input, $0.66/1M output; peak: $0.014, $0.44, and $1.32 respectively

Commercial

🚀Function Overview

An experimental hosted model for screenshot reading, chart analysis, visual agents, and mixed text-image workflows. Try it now and see the results!

Key Features

Separate request model ID deepseek-v4-flash-vision-exp
Mixed text and image input with text output
1M token context length and 384K maximum output
JPEG, PNG, GIF, and WebP input through base64, external URLs, or the Files API
Chat Completions, Responses API, and Anthropic-compatible Messages support
JSON output, tool calls, thinking and non-thinking modes
Up to 384 billed visual tokens per image after automatic resizing
Off-peak pricing of $0.007 cache-hit input, $0.22 cache-miss input, and $0.66 output per 1M tokens
Peak pricing of $0.014 cache-hit input, $0.44 cache-miss input, and $1.32 output per 1M tokens
Experimental API release without published weights, architecture details, or a model card

Use Cases

•Read screenshots, forms, charts, and visual interface state
•Build tool-using agents that combine image understanding with text instructions
•Reuse uploaded images through the Files API
•Test DeepSeek multimodal workflows before choosing a production model
•Compare image-token cost and compatibility across OpenAI- and Anthropic-format clients

⚙️Input Parameters

messages

array

Text and image content blocks. Images are accepted only in user messages for Chat Completions and Anthropic-format Messages.

image

JPEG, PNG, GIF, or WebP supplied as base64, an external URL, or a reusable Files API file_id.

detail

string

Optional image processing level. `low` downsizes to 512 by 512; high, original, and auto currently preserve original detail before automatic tokenization limits.

tools

array

Optional function definitions for tool-using multimodal workflows.

💡Usage Examples

Example 1

Input Parameters

{
  "model": "deepseek-v4-flash-vision-exp",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Describe the error shown in this screenshot and suggest the next diagnostic step."
        },
        {
          "type": "image_url",
          "image_url": {
            "url": "https://example.com/error.png",
            "detail": "low"
          }
        }
      ]
    }
  ]
}

Output Results

A text response grounded in the supplied image. Verify client support for multimodal content blocks before production use.

Quick Actions

Technical Specifications

Hardware Type: DeepSeek API
Commercial Use: Supported
Pricing: Off-peak: $0.007/1M cache-hit input, $0.22/1M cache-miss input, $0.66/1M output; peak: $0.014, $0.44, and $1.32 respectively

Related Keywords

DeepSeek V4 Flash Vision ExpDeepSeek-V4-Flash-Vision-Expdeepseek-v4-flash-vision-expDeepSeek vision APIDeepSeek image inputDeepSeek multimodal modelDeepSeek Files APIDeepSeek vision pricing