Home / Technology / China's AI Leap: New Models Challenge US Giants
China's AI Leap: New Models Challenge US Giants
24 Apr
Summary
- DeepSeek unveiled new AI models, challenging US tech rivals.
- Pro version leads open models in math and coding performance.
- Previous model release caused privacy and security concerns.
A year after its initial breakthrough, China's DeepSeek has introduced the latest iterations of its artificial intelligence chatbots, DeepSeek-V4-Pro and DeepSeek-V4-Flash. These new preview versions are positioned to compete directly with leading US AI developers such as OpenAI and Google. The company highlighted that its "pro" model surpasses all other open models in mathematical and coding benchmarks, and is second only to Google's Gemini 3.1-Pro in global knowledge.
The "flash" variant provides comparable reasoning abilities to the "pro" model but is designed for quicker responses and reduced operational costs. This development echoes the impact of DeepSeek-R1, released in January of the previous year, which demonstrated capabilities rivaling ChatGPT and Gemini. Its emergence was notably lauded by venture capitalist Marc Andreessen as an "AI's Sputnik moment."
DeepSeek's prior model attracted attention due to its claimed development costs of less than $6 million, a stark contrast to the substantial budgets typical in Silicon Valley. However, the company's advancements have also led to scrutiny and, in several countries including multiple US states, Australia, Taiwan, South Korea, Denmark, and Italy, resulted in bans or restrictions citing data protection, privacy, and national security concerns.
