China’s DeepSeek unveils new AI model, one year after disrupting US rivals
▼ Summary
– DeepSeek released a preview of its next-generation AI model V4, claiming it competes with leading closed-source US systems from Anthropic, Google, and OpenAI.
– The V4 model shows major improvement in coding, a key capability for AI agents, and marks a milestone for China’s chip industry due to compatibility with Huawei technology.
– The release follows DeepSeek’s R1 model from a year ago, which rattled the US AI industry by claiming lower training costs than US systems.
– DeepSeek has not disclosed V4’s training costs or hardware, while US officials have accused the company of using banned Nvidia chips.
– Anthropic claims DeepSeek misused its Claude model to improve its own products.
A year after shaking up the US artificial intelligence sector, Chinese AI firm DeepSeek has unveiled a preview of its eagerly awaited next-generation model, V4. The company asserts that this open-source system can go head-to-head with leading closed-source competitors from Anthropic, Google, and OpenAI.
According to DeepSeek, V4 represents a substantial leap forward compared to earlier versions, particularly in coding capabilities. This area has become a cornerstone for AI agents and has fueled the success of tools such as ChatGPT Codex and Claude Code. The launch also signals a notable achievement for China’s semiconductor industry, as DeepSeek has emphasized the model’s compatibility with domestic Huawei technology.
This release arrives exactly one year after DeepSeek stunned US tech giants with R1, a model it claimed was developed at a fraction of the training cost of top American systems. For V4, however, the company has not disclosed the training costs or the specific hardware used. Meanwhile, US officials have alleged that DeepSeek utilized restricted Nvidia chips, and Anthropic has accused the firm of improperly using Claude to enhance its own models.
(Source: The Verge)
