Lead

Chinese startup DeepSeek released its next-generation artificial intelligence model on Friday, touting "drastically reduced" compute and memory costs and an ultra-long context window of one million words. The release, more than a year after the company stunned the tech world with its low-cost R1 reasoning model, comes amid heightened US-China rivalry over AI, with the White House on Thursday accusing Chinese entities of a massive effort to steal AI technology.

Coverage Comparison

Coverage of the release from France 24 (via AFP) and the South China Morning Post broadly aligns on the core facts: DeepSeek unveiled V4 in two versions, V4-Pro and V4-Flash, with the Pro version boasting 1.6 trillion parameters and the Flash version 284 billion. Both models feature a context window of one million tokens, a key capability that determines how much information an AI system can process at once.

However, the two outlets frame the story differently. France 24's report places the release squarely in the context of US-China technological rivalry, noting the White House's accusation that Chinese entities have engaged in technology theft. The South China Morning Post, by contrast, focuses on the technical specifications and the competitive positioning of the model against leading US closed-source models from OpenAI and Google DeepMind, without referencing the political backdrop.

Neither outlet suggests any disagreement on the factual details; the divergence is in emphasis and context.

Key Claims

  • DeepSeek released V4, an open-source foundational AI model, in two versions: V4-Pro with 1.6 trillion parameters and V4-Flash with 284 billion parameters. This is reported by both France 24 and the South China Morning Post.
  • The model features a context window of one million tokens, which DeepSeek claims is achieved with "world-leading" cost efficiency. This is reported by both sources.
  • The model has been optimized for popular AI agent products such as Claude Code, OpenClaw, OpenCode, and CodeBuddy, according to DeepSeek's statements as carried by France 24.
  • The White House accused Chinese entities of stealing AI technology, a claim reported by France 24 and confirmed by the White House's own statements.
  • DeepSeek's AI tools have been widely adopted by Chinese municipalities, healthcare institutions, the financial sector, and other businesses. This claim appears in a single source and has not been independently verified.

Perspectives

DeepSeek

DeepSeek positions the V4 release as a leap forward in efficiency and capability, emphasizing the "drastically reduced" costs and the one-million-word context window, which it says achieves leadership in both domestic and open-source fields. The company frames the release as a democratizing force for AI, making long-context processing accessible beyond high-end research labs.

US Government

The White House has accused Chinese entities of a massive effort to steal AI technology, reflecting a broader US concern about China's rapid advances in AI. This accusation frames DeepSeek's success within a narrative of strategic competition and potential intellectual property theft.

Industry Analysts

Tech research firm iiMedia founder Zhang Yi calls the release a "genuine inflection point" for the industry, noting it addresses long-standing issues of slower performance and higher costs associated with long context lengths. He anticipates that ultra-long context support could become a standard feature, enabling mainstream commercial applications.

South China Morning Post

While not explicitly offering a distinct perspective, the SCMP's coverage emphasizes the competitive aspect, noting that DeepSeek's V4 rivals leading US closed-source models. The outlet's technical focus suggests a view of the release as a significant milestone in AI development, independent of political tensions.

This article was compiled from reporting by France 24 (AFP) and the South China Morning Post.