On August 6, ByteDance CEO Liang Rubo highlighted at the 2026 mid-year all-staff meeting that the company's large language models (LLMs) currently lag behind leading overseas counterparts. Nevertheless, ByteDance is committed to continuing independent research, strengthening its foundational capabilities, accepting short-term setbacks, and focusing on continuous optimization for long-term growth. Liang emphasized that this commitment to independent research is not a response to external pressures but a strategic choice aimed at fostering a culture of delayed gratification within the team and laying a robust foundation for achieving Artificial General Intelligence (AGI) in the long run.

Tan Dai, the head of ByteDance's creativity service platform, pointed out that in the business-to-business (ToB) sector, only state-of-the-art (SOTA) models can initially unlock scenarios with high economic value. He noted that the video generation model Seedance 2.0 has successfully transitioned this technology from entertainment applications to serious commercial content production. Moreover, its usage volume is notably higher on weekdays compared to weekends, indicating its growing adoption in professional settings.