China’s DeepSeek Unveils V4 AI Model to Rival US Competitors
Updated
Updated · South China Morning Post · Apr 24
China’s DeepSeek Unveils V4 AI Model to Rival US Competitors
54 articles · Updated · South China Morning Post · Apr 24
Chinese AI startup DeepSeek has released a preview of its V4 model, claiming performance competitive with leading US models from OpenAI and Google.
The open-source V4 comes in 'pro' and 'flash' versions, boasting a 1 million token context window and significant improvements in reasoning and agentic abilities.
Backed by Huawei’s domestic chips, V4 reflects China’s push for AI self-reliance amid US export controls, but faces ongoing allegations of model distillation from US firms.
How might DeepSeek’s open-source strategy impact the global AI landscape and the competitive advantage of proprietary models?
How will DeepSeek’s collaboration with Huawei and Cambricon influence future AI chip development and deployment in China?
What independent evaluations are planned to verify DeepSeek V4's performance and cost efficiency against U.S. models?
Will DeepSeek V4's one million token context window reshape how businesses and researchers use AI for large-scale reasoning tasks?
How credible are DeepSeek’s claims of outperforming GPT-5.2 and Gemini 3.0-Pro in reasoning and agentic benchmarks?
Could accusations of model distillation lead to stricter regulations or new forms of AI governance between China and the U.S.?