DeepSeek V4 enters the AI market as an open-source rival

Chinese artificial intelligence company DeepSeek has launched a preview of its latest model, DeepSeek V4, positioning itself as a major competitor to Western tech giants. Unlike the closed systems from OpenAI and Anthropic, this new offering is an open-source model available for public modification under an MIT license. The release highlights China's growing influence in the AI landscape by providing high-performance capabilities at a fraction of the cost of leading U.S. models.

According to Mashable, DeepSeek has officially unveiled its V4 preview, marking a significant milestone in the ongoing global competition for artificial intelligence supremacy. While companies like OpenAI, Google DeepMind, and Anthropic maintain proprietary control over their frontier models, DeepSeek distinguishes itself by offering its technology as an open-source resource that anyone can download and modify.

Model Variations and Technical Specifications

The new release includes two distinct versions tailored for different use cases. The DeepSeek-V4-Pro is a massive model featuring 1.6-trillion parameters, while the DeepSeek-V4-Flash offers a more streamlined architecture with a total of 284-billion parameters. These models are designed to be seamlessly integrated with prominent AI agents such as Claude Code, OpenClaw, and OpenCode.

While DeepSeek V4 currently sits slightly behind the top frontier models on popular leaderboards like Arena and Artificial Analysis, the company claims significant strides in coding and agentic tasks. The full system card for these models is currently accessible via Hugging Face for developers to review technical details.

Cost Efficiency and Market Positioning

One of the most striking aspects of DeepSeek V4 is its aggressive pricing strategy compared to its American counterparts. Following the success of the R1 model in early 2025, which demonstrated that powerful AI could be built efficiently, DeepSeek continues to undercut the market on price points.

The API pricing for DeepSeek V4 stands at $1.74 per 1 million input tokens and $3.48 per 1 million output tokens for a 1 million context window. This represents a substantial discount compared to other major players:

GPT-5.5: $5 per 1 million input tokens and $30 per 1 million output tokens

Claude Opus 4.7: $5 per 1 million input tokens and $25 per 1 million output tokens

Google Gemini 3.1 Pro: $2 per 1 million input tokens and $12 per 1 million output tokens

By offering performance that rivals the latest frontier models at roughly one-sixth of the cost, DeepSeek is positioning itself to gain rapid adoption among developers who prioritize economy. This move underscores a shift toward making high-tier AI capabilities more accessible and affordable for global users.

FAQ

What are the different versions of DeepSeek V4?

DeepSeek released two distinct versions: DeepSeek-V4-Pro, which features 1.6 trillion parameters, and DeepSeek-V4-Flash, which has a more streamlined architecture with 284 billion parameters. Both models are designed for integration with AI agents like Claude Code and OpenClaw.

How does DeepSeek V4 pricing compare to other models?

DeepSeek V4 costs $1.74 per 1 million input tokens and $3.48 per 1 million output tokens. This is significantly cheaper than GPT-5.5, which costs $5 for input and $30 for output, and Claude Opus 4.7, which costs $5 for input and $25 for output.

Where can developers find the technical details for DeepSeek V4?

The full system card for the DeepSeek V4 models is currently accessible via Hugging Face for developers to review all technical specifications.

Model Variations and Technical Specifications

Cost Efficiency and Market Positioning

FAQ

Fresh news on our Telegram