Models·DeepSeek

DeepSeek V4 Flash Vision Exp

deepseek/deepseek-v4-flash-vision-exp

An experimental DeepSeek model that reads images and text and responds in text.

$0.22Input, per million tokens
$0.65Output, per million tokens
1.05MContext window
262KLongest answer

1.05M tokens of context — about 786,432 words, on a scale that ends at 2M.

What it can do

  • Reasoning
  • Vision
  • Tool use
  • Open weights
  • Takes text
  • Takes image
  • Gives text

Released 21 Aug 2026 · checked 3 Oct 2026.

About it

DeepSeek V4 Flash Vision Exp is an experimental model from DeepSeek with image understanding alongside text skills. It can reason and use tools, and its weights are open. It accepts text and images but returns text. The model has a large context window and a maximum output of 262,144 tokens. Its listed input and output prices are charged per million tokens.

Review

In a line

DeepSeek V4 Flash Vision Exp is an experimental model from DeepSeek that can understand images as well as text. It returns answers in text.

What it's good at

It suits tasks that combine text and images, and it can handle reasoning and tool use. Its text capabilities match those of the base model, DeepSeek V4 Flash 0731.

Its weights are open, and it supports image input.

Price in plain terms

Input costs $0.2156 per million tokens, and output costs $0.6468 per million tokens. A page of about 500 words costs roughly 0.014 cents to read. One chat turn with about 1,000 input tokens and 500 output tokens costs about 0.054 cents; 1,000 such turns cost about $0.54.

Context and limits

It can take in up to 1,048,576 tokens at once, which is the amount of text and image-related input it can consider together. It can write up to 262,144 tokens in one answer.

It accepts text and images, but its output is text.

Ways to use it

You can use the model through providers' APIs or apps and chat tools that offer it. Its model ID is deepseek/deepseek-v4-flash-vision-exp.

Explore DeepSeek models, vision language models, and open-weight models.

Verdict

Consider it if you need an open-weight model that can work with both images and text, reason, and use tools. Its token prices are low in the supplied examples, including about $0.54 for 1,000 typical chat turns. It may not fit if you need image output, since it responds in text, or if you want a non-experimental model.

Verdict

DeepSeek V4 Flash Vision Exp is worth considering for people comparing models that can take images as well as text, especially when open weights, reasoning, and tool use matter. Its example costs are small for page reading and chat turns, though actual spending depends on token use. Look elsewhere if you need answers in image form or prefer a model that is not labeled experimental.

  • Accepts both text and images.
  • Supports reasoning and tools.
  • Open weights are available.
  • Large context window and high maximum output.
  • Example costs include about $0.54 per 1,000 chat turns.
  • Experimental version.
  • Returns text only.
  • Use is through providers' APIs, apps, or chat tools that offer it.
  • Pick it ifPeople comparing open-weight models with image input.; Tasks that combine images, text, reasoning, or tools.
  • Skip it ifYou need the model to generate images.; You want a model that is not experimental.

Pricing

The listed rates are $0.2156 per million input tokens and $0.6468 per million output tokens. At the supplied usage examples, reading a page costs about 0.014 cents, one chat turn about 0.054 cents, and 1,000 turns about $0.54.

Questions

Can DeepSeek V4 Flash Vision Exp understand images?

Yes. It accepts image and text input, and returns text.

Can it use tools or reason through a task?

Yes. The model supports tool use and reasoning.

How much does it cost?

Input is $0.2156 per million tokens and output is $0.6468 per million tokens. The examples put 1,000 chat turns at about $0.54.

How much can it read at once?

Its context window is 1,048,576 tokens.

How long can its answer be?

Its maximum output is 262,144 tokens.

How can I use it?

Through providers' APIs or apps and chat tools that offer the model. Its model ID is deepseek/deepseek-v4-flash-vision-exp.

Are its weights open?

Yes. Its weights are listed as open.

On these boards

More from DeepSeek