ByteDance reportedly targets mega AI model to match Mythos scale

Chinese tech giant ByteDance is reportedly training a massive artificial intelligence model, aiming for a scale of up to 10 trillion parameters. This ambitious project, if realized, could position ByteDance's AI system to rival Anthropic's advanced Mythos model, according to a report by the Financial Times citing sources familiar with the matter.

With 10 trillion parameters, ByteDance's potential model would be more than three times larger than China's current leading AI, Moonshot AI's Kimi K3, which boasts 2.8 trillion parameters. Parameters in AI models are numerical settings learned from data, often used as a benchmark for scale, though not necessarily a direct indicator of capability. Before Kimi K3's release, models like Meituan's LongCat-2.0 and DeepSeek's V4-Pro, with 1.6 trillion parameters, led China's AI sector.

Direct comparisons with top U.S. AI models remain challenging, as companies like Anthropic and OpenAI do not disclose parameter counts for their flagship systems, including Fable, Mythos, or GPT-5.5. Industry estimates suggest Anthropic's Mythos 5 has around 8 trillion parameters, placing ByteDance's new model in close proximity in terms of scale.

This development underscores the accelerating pace of AI innovation among Chinese tech firms. They are striving to keep up with global competitors, particularly U.S. rivals, in the race to develop more powerful AI systems without incurring prohibitive operational costs. The ByteDance model is currently in its pre-training phase, a process typically lasting three to six months before it can be fine-tuned and released.