Recently, reports have emerged that ByteDance founder Zhang Yiming said at a recent internal all-hands meeting for AI teams that ByteDance is capable of competing with OpenAI and Anthropic, but that truly catching up to the international technological frontier may not happen until 2029 or 2030 at the earliest.
He emphasized that ByteDance's large models should benchmark against the world's top tier, and that compared with top overseas AI companies, the current Seed series models still lag in the scale of their pre-training datasets—an issue that requires close attention at this stage.
According to foreign media reports, ByteDance's Seed team is pre-training a model with up to 10 trillion parameters, a scale expected to reach the level of Anthropic's most advanced models.
This is not the first time Zhang Yiming has emphasized a long-term path for ByteDance's AI development. Previously, he said at an internal meeting that model R&D must adhere to long-termism and delayed gratification, and that it should not rely on the outputs of other models to win short-term leaderboard rankings.
These remarks once again clarify ByteDance's competitive goals in large-model R&D, and also reflect its emphasis on foundation model capabilities, pre-training data scale, and long-term technological accumulation.
