DeepSeek rolls out flagship AI model, takes on OpenAI and Anthropic
DeepSeek unveiled preview versions of its new flagship AI model, V4 Flash and V4 Pro, claiming significant improvements in reasoning and coding benchmarks. The Chinese startup has been accused by US authorities and tech executives of using illegal technology and methods to develop its models.
DeepSeek, a Hangzhou-based startup, has introduced the V4 Flash and V4 Pro preview series of its AI model, claiming significant improvements in reasoning and agentic activities. The new model can function with a million-token context length and features a Hybrid Attention Architecture. DeepSeek's previous model, R1, sparked a trillion-dollar stock market selloff and was constructed at a fraction of the cost of state-of-the-art AI systems. US tech companies are expected to spend almost $650 billion on data centres and AI infrastructure by 2026. DeepSeek has faced scrutiny over its development methods, with accusations of using 'distillation' and accessing prohibited Nvidia Corp. chips. The startup's actions have caused a frenzy in China, with competitors rushing to update their models.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.