DeepSeek releases V4 AI model with lower costs but fails to close gap with U.S. leaders like OpenAI

DeepSeek, China's leading AI company, unveiled preview versions of V4, its new flagship model, more than a year after releasing R1 in January 2024. V4 costs less to use than competing models but does not meaningfully narrow the performance gap with frontier U.S. AI systems.

Unlike some feared when R1 launched, V4 is not competitive with top-tier American models in core AI capabilities (how well it understands tasks, reasons through complex problems, and generates accurate answers). The gap persists despite DeepSeek's cost advantage, meaning cheaper pricing alone does not translate to technical leadership.

For business professionals and students choosing AI tools: U.S.-based systems (like those from OpenAI) remain the safest bet if you need cutting-edge performance, though DeepSeek's lower operating costs may appeal to cost-sensitive projects where top-tier performance is less critical. For U.S. policymakers, this outcome reduces immediate competitive pressure around AI dominance.