DeepSeek V4 Officially Launched, Official API Pricing Announced
Keywords: deepseek v4, deepseek official website, deepseek tutorial, deepseek v4 price
Publication Date: April 25, 2026
Author: DeepSeek HK
Today, DeepSeek officially announced the API pricing plan for the new V4 series models, marking the official commercial launch of this highly anticipated next-generation large model for developers worldwide. The DeepSeek V4 series includes two versions: Flash and Pro, both supporting 1 million token ultra-long context. While performance has been greatly improved, cost optimization has been achieved, providing a more competitive choice for AI application development.
Model and Pricing Description
DeepSeek API pricing is calculated per million tokens. A token is the smallest unit of text recognized by the model, which can be a word, number, or even punctuation. Fees are calculated based on the total number of input and output tokens actually processed by the model.
Detailed Model Parameter Comparison
| Feature | deepseek-v4-flash* | deepseek-v4-pro |
|---|---|---|
| BASE URL (OpenAI Format) | https://api.deepseek.com | https://api.deepseek.com |
| BASE URL (Anthropic Format) | https://api.deepseek.com/anthropic | https://api.deepseek.com/anthropic |
| Model Version | DeepSeek-V4-Flash | DeepSeek-V4-Pro |
| Thinking Mode | Supports both non-thinking and thinking (default) modes. See Thinking Mode for switching instructions | Supports both non-thinking and thinking (default) modes. See Thinking Mode for switching instructions |
| Context Length | 1M | 1M |
| Maximum Output Length | Up to 384K | Up to 384K |
| JSON Output | ✓ | ✓ |
| Tool Calls | ✓ | ✓ |
| Chat Prefix Completion (Beta) | ✓ | ✓ |
| FIM Completion (Beta) | Non-thinking mode only | Non-thinking mode only |
| 1M Input Tokens (Cache Hit) | $0.028 | $0.145 |
| 1M Input Tokens (Cache Miss) | $0.14 | $1.74 |
| 1M Output Tokens | $0.28 | $3.48 |
- Note: The model names
deepseek-chatanddeepseek-reasonerwill be gradually phased out in the future. For compatibility, these two names currently correspond to the non-thinking mode and thinking mode ofdeepseek-v4-flashrespectively.
Fee Deduction Rules
Actual consumption amount = token usage × corresponding unit price. Fees will be deducted directly from your recharge balance or gift balance. When both balances exist, the system will prioritize the use of the gift balance.
Product prices may be adjusted according to market conditions, and DeepSeek reserves the final right to adjust prices. Developers are advised to recharge according to actual usage needs and regularly check the official pricing page for the latest information.
Core Advantage Analysis
While maintaining high performance, the DeepSeek V4 series implements a highly competitive pricing strategy, which is especially suitable for various AI application scenarios:
Cost-effective Computing Power Selection
The deepseek-v4-flash version provides excellent performance at an extremely low price, especially suitable for scenarios with high throughput and high response speed requirements, such as intelligent customer service, content generation, regular chat applications, etc. Its price of only $0.028 per million input tokens when cache hits significantly reduces the operating cost of large-scale applications.
Flagship Performance Experience
The deepseek-v4-pro version is optimized for complex reasoning and professional field tasks, and performs better in scenarios such as mathematical calculation, code generation, and professional knowledge Q&A, suitable for professional application scenarios that require high-precision output.
Ultra-long Context Support
Both versions support 1 million token ultra-long context and maximum 384K output length, which can handle super large text tasks such as entire books, long documents, and complete project code, providing a strong foundation for developing complex AI applications.
Quick Start Guide
Getting started with DeepSeek V4 is very simple:
- Visit the DeepSeek official website to register an account
- Obtain the API key
- Make interface calls according to the official documentation
- Start building your AI application
The release of DeepSeek V4 provides developers with more choices. Whether you are an individual developer or an enterprise-level application, you can find a computing power solution suitable for your needs in this model system. With the continuous improvement of the model ecosystem, we believe that DeepSeek V4 will become an important cornerstone of AI application development, promoting the birth of more innovative products.