ByteDance is training a massive AI model with up to 10 trillion parameters, placing it near the estimated scale of Anthropic's Mythos 5. To support its expanding AI ambitions, ByteDance has established a new top-level AI data and safety department led by former TikTok executive Wang Yinglei. This organizational shift comes as Chinese tech firms accelerate model releases to compete with U.S. rivals like OpenAI and Anthropic in a rapidly intensifying global AI race.
ByteDance 10 trillion parameter model
- ▪The 10 trillion parameter artificial intelligence model being trained by ByteDance is currently undergoing pre-training, a process that typically lasts three to six months, according to a Financial Times report on August 7, 2026.
- ▪ByteDance is training an artificial intelligence model with as many as 10 trillion parameters, according to a Financial Times report on August 7, 2026.
AI data department formation
- ▪ByteDance has formed a new top-level department focused on artificial intelligence data and safety, placing it alongside Seed, Flow, and Douyin within the company's organizational structure, according to a TechNode report on August 12, 2026.
- ▪The newly formed ByteDance artificial intelligence data and safety department grew out of a global data team established in 2023, according to a TechNode report on August 12, 2026.
- ▪The ByteDance artificial intelligence data and safety department's work spans data sourcing, synthetic-data generation, data cleaning, standards, and quality evaluation for ByteDance's foundation models, according to a TechNode report on August 12, 2026.
AI safety department formation
- ▪The team that became ByteDance's artificial intelligence data and safety department previously supported TikTok, Dola (the overseas version of Doubao), and Seed, according to a TechNode report on August 12, 2026.
- ▪The newly formed ByteDance artificial intelligence data and safety department is led by Wang Yinglei, a former TikTok executive who oversaw platform responsibility and livestreaming, according to a TechNode report on August 12, 2026.
China-US AI competition
- ▪Direct comparisons between Chinese and United States artificial intelligence models are difficult because United States companies such as Anthropic and OpenAI do not disclose parameter counts for systems like Fable, Mythos, or GPT-5.5, as of August 2026.
- ▪Chinese tech firms are accelerating their artificial intelligence model release cycles to keep up with the global AI race and compete with United States rivals to build powerful systems without making them prohibitively expensive to run, as of August 2026.
AI model parameter comparisons
- ▪A 10 trillion parameter artificial intelligence model would be more than three times the size of Chinese startup Moonshot AI's Kimi K3 model, which has 2.8 trillion parameters, as of August 2026.
- ▪Prior to the release of Moonshot AI's Kimi K3, Meituan's LongCat-2.0 and DeepSeek's V4-Pro led China's artificial intelligence industry with 1.6 trillion total parameters, as of August 2026.
- ▪Industry estimates cited by the Financial Times indicate that Anthropic's Mythos 5 has about 8 trillion parameters and Fable 5 has about 5 trillion parameters, as of August 2026.
Story comments
Loading comments…