Models·DeepSeek

DeepSeek V4 Flash 0731 Review

deepseek/deepseek-v4-flash-0731

DeepSeek V4 Flash 0731 is an open-weight text model for coding, reasoning, and agent workflows.

$0.011Input, per million tokens
$1.28Output, per million tokens
1.05MContext window
944KLongest answer

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

What it can do

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

Released 31 Jul 2026 · checked 3 Oct 2026.

About it

DeepSeek V4 Flash 0731 is a DeepSeek text model released July 31, 2026. It uses a sparse mixture-of-experts design, with 13 billion active parameters from 284 billion total. The model supports reasoning and tools, accepts text, and returns text. Open weights are available, with a context window of 1,048,576 tokens.

Review

In a line

DeepSeek V4 Flash 0731 is a text model from DeepSeek, released July 31, 2026. It is a retrained revision in the DeepSeek family, with open weights available.

Its sparse mixture-of-experts design uses 13 billion active parameters out of 284 billion total.

What it's good at

The model is suited to coding, reasoning, and agent workflows. It supports reasoning and tool use, so software can ask it to work with tools during a task.

It handles text input and produces text output.

Price in plain terms

Input costs $0.0115 per million tokens. Output costs $1.28 per million tokens. A token is a small piece of text counted for billing.

Reading a 500-word page costs under 0.01 cents in input charges. A chat turn with about 1,000 tokens in and 500 out costs about 0.065 cents. One thousand such turns cost about $0.65.

For comparisons with other models, see cheapest language models.

Context and limits

The model can take up to 1,048,576 tokens in one context. That is the amount of text it can consider at once.

A single response can contain up to 943,718 tokens.

Its 13 billion active parameters are the portion used for each request; the full model has 284 billion parameters.

Ways to use it

Use it for coding, reasoning, or agent tasks in software that supports tool-enabled models. Its tool support lets an application give it actions to perform during a workflow.

Because the weights are open, people who run models themselves can work with the published files. The model is identified as deepseek/deepseek-v4-flash-0731, with the Hugging Face ID deepseek-ai/DeepSeek-V4-Flash-0731.

You can compare it with other DeepSeek models or open-weight models.

Verdict

Choose this model if you need a text system for programming, step-by-step reasoning, or tool-driven agent work and want open weights. Its very large context may suit tasks that involve a lot of text.

Look elsewhere if your application needs images, since the listed input and output types are text only. Also check the separate input and output rates when estimating costs, because generated text is priced much higher than supplied text.

Verdict

DeepSeek V4 Flash 0731 fits builders who want open model weights, long text handling, and support for reasoning or tool-based tasks. Its low input rate keeps document-heavy requests inexpensive, while generated text carries a higher charge. The model is not a match for image-based work because it is text-only. Before adopting it, confirm that your software can use the model identifier or run the published weights, and estimate spending from both sides of the token bill.

  • Open weights are available.
  • Supports reasoning and tool use.
  • Handles up to 1,048,576 context tokens.
  • Designed for coding and agent workflows.
  • Low input cost per million tokens.
  • Text input and output only.
  • Output costs $1.28 per million tokens.
  • Maximum response is 943,718 tokens.
  • No specific app integrations are listed.
  • Pick it ifCoding and reasoning tasks; Tool-based agent workflows; Long text contexts
  • Skip it ifYou need image input or image output.; Your software cannot run open weights or use this model ID.; Generated-text costs need to be lower.

Pricing

Input is $0.0115 per million tokens and output is $1.28 per million tokens. A 500-word page costs under 0.01 cents to read; 1,000 typical chat turns cost about $0.65.

Questions

What is DeepSeek V4 Flash 0731?

It is a DeepSeek text model released July 31, 2026, for coding, reasoning, and agent workflows.

Is DeepSeek V4 Flash 0731 open source?

Its weights are open. The listed Hugging Face ID is deepseek-ai/DeepSeek-V4-Flash-0731.

How much does it cost?

Input costs $0.0115 per million tokens. Output costs $1.28 per million tokens.

How much text can it read?

Its context window is 1,048,576 tokens.

How long can one answer be?

The maximum output is 943,718 tokens.

Can it work with tools?

Yes. Tool support is listed, alongside reasoning support.

Does it handle images?

No image capability is listed. Inputs and outputs are text.

On these boards

More from DeepSeek