According to a report by The Paper, ByteDance founder Zhang Yiming recently had an open discussion about the AI business during an internal sharing session and provided a timeline. In a recent all-hands meeting for the AI team, Zhang stated that ByteDance has the capability to compete with OpenAI and Anthropic, but to truly catch up to cutting-edge technology, it will take "at least until 2029 or 2030." He also made it clear that ByteDance's large models will be benchmarked against the international top tier.
Regarding the existing gap, Zhang did not shy away from the issue. He admitted that compared to leading overseas AI companies, the pre-training dataset scale of ByteDance's Seed series models is still relatively small, which is currently the most pressing concern for ByteDance AI. In the composition of model capabilities, data scale is often the hardest aspect to quickly catch up on: computing power can be purchased, teams can be recruited, but accumulating high-quality corpora requires time and sustained investment, which cannot be solved by simply increasing volume in the short term.
ByteDance Plans to Catch Up to AI Technology Frontiers by 2029 or 2030
However, ByteDance's actions are not slow. Foreign media reports indicate that the ByteDance Seed team is pre-training a model with a parameter scale of up to 100 trillion, which can rival the most advanced models from Anthropic. If this claim is true, it suggests that ByteDance is already preparing according to the standards of the top tier in terms of scale.
This is not the first time Zhang Yiming has discussed the development path of AI internally. Previously, The Paper's reporters learned from insiders at ByteDance that he had stated in an internal meeting that model development should adhere to long-termism and delayed gratification, rather than merely using others' outputs for short-term ranking.

