ByteDance has unveiled its latest AI model, Doubao-1.5-pro, which is reported to match the performance benchmarks of OpenAI's GPT-4o and outperform other models such as deepseek-v3 and llama3.1-405B on popular benchmarks. The model features a 'Deep Thinking' mode that surpasses previous iterations on the AIME benchmark. Built on a mixture of experts (MoE) architecture, Doubao-1.5-pro operates with significantly fewer activated parameters compared to its competitors, making it a cost-efficient alternative in the escalating AI competition between China and the U.S. This launch is part of ByteDance's broader strategy to position itself as a leader in artificial general intelligence (AGI). The introduction of Doubao-1.5-pro has raised concerns in Silicon Valley, as it demonstrates the potential for Chinese AI models to rival those developed in the U.S. despite using less powerful chips.