China's AI startup DeepSeek has unveiled its latest large language model, DeepSeek-V4, claiming it outperforms OpenAI's GPT-4 in reasoning and coding tasks while being significantly cheaper to operate. The model reportedly achieves 92% of GPT-4's performance at just 20% of the computational cost, potentially disrupting the global AI market. DeepSeek's CEO stated the breakthrough comes from novel neural architecture optimizations and proprietary training techniques. The release comes as China intensifies efforts to compete with US AI leaders, with DeepSeek having raised over $500 million in funding this year. Industry analysts note the model's cost efficiency could accelerate AI adoption in price-sensitive markets.