DeepSeek releases V4 AI model matching top U.S. competitors at 1/6th the cost, forcing a price collapse in enterprise AI

DeepSeek, a Chinese AI startup, released DeepSeek-V4 last night—a 1.6-trillion-parameter AI model (a measure of the AI's size and capability) available free under an open-source MIT license. The model performs near or better than Claude Opus 4.7 and GPT-5.5, the most advanced closed-source AI systems from Anthropic and OpenAI, but costs about 1/6th as much to run via API.

Unlike closed-source competitors, DeepSeek-V4 uses a Mixture-of-Experts architecture (a technique that activates only the parts of the AI needed for each task), which lets it match the reasoning ability of much larger systems while using fewer computing resources. This means businesses pay significantly less per AI query without waiting longer for answers.

For business professionals and students, this forces a pricing reset: AI chat APIs and enterprise AI tools will likely drop prices to compete, making AI assistance cheaper to integrate into workflows. For software developers, the free open-source license means they can build AI features into products without licensing fees from U.S. companies.