Three Reasons Why DeepSeek's New V4 Model Matters

Chinese AI firm DeepSeek has released a preview of V4, its long-awaited new flagship model. This marks the company's most significant release since its R1 reasoning model debuted in January 2025. The new V4 model can process much longer prompts than its previous generation, utilizing a new design that handles large volumes of text more efficiently.

Reason 1: Breaking New Ground for Open-Source AI

Like its predecessor, DeepSeek V4 is open source, allowing developers and companies to download, use, and modify the technology. This release provides access to frontier AI capabilities without the burden of skyrocketing costs. It is available in two versions on DeepSeek's website, app, and API: V4-Pro, designed for coding and complex agent tasks, and V4-Flash, built to be faster and cheaper to run. Both versions feature reasoning modes that display each step of the problem-solving process.

Reason 2: Unrivaled Cost Efficiency and Performance

For V4-Pro, DeepSeek charges $1.74 per million input tokens and $3.48 per million output tokens, which is a fraction of the cost of comparable models from OpenAI and Anthropic. V4-Flash is even cheaper, at $0.14 per million input tokens and $0.28 per million output tokens. Despite the low cost, V4-Pro matches the performance of leading closed-source models, including Anthropic's Claude-Opus-4.6, OpenAI's GPT-5.4, and Google's Gemini-3.1. It also exceeds open-source rivals Alibaba's Qwen-3.5 and Z.ai's GLM-5.1 on coding, math, and STEM. An internal survey of 85 experienced developers revealed that over 90 percent included V4-Pro as a top choice for coding tasks, and the model is optimized for popular agent frameworks like Claude Code, OpenClaw, and CodeBuddy.

Reason 3: A Major Win for Chinese Chipmakers

The long-awaited V4 model is highly efficient, processing massive amounts of text with a streamlined design. By achieving frontier-level performance on highly optimized architecture, the release represents a massive win for Chinese chipmakers, demonstrating that cutting-edge AI can be trained and run efficiently without relying on infinite hardware resources.