Reports indicate ByteDance is developing an AI model that could potentially surpass 'Claude Mythos 5'.

The Financial Times reported that 'Chinese company ByteDance is developing an AI model with up to 10 trillion parameters.'

ByteDance targets mega AI model nearing Anthropic's Mythos

ByteDance targets mega AI model that could match Mythos scale, FT reports | Reuters
https://www.reuters.com/technology/bytedance-targets-mega-ai-model-nearing-anthropics-mythos-ft-reports-2026-08-07/

According to the Financial Times, citing sources, ByteDance is in the early stages of training a model with up to 10 trillion parameters, a stage that typically takes 3 to 6 months. After that, if everything goes smoothly, the model will be fine-tuned and released to the public.

The number of parameters indicates the size of an AI model; a larger number suggests more knowledge. However, a large number of parameters doesn't necessarily translate directly to high processing power. Nevertheless, considering that a model with 2.4 trillion parameters has just emerged in China, it's clear that 10 trillion parameters represents an unimaginable scale.

China's Alibaba announces 'Qwen3.8-Max,' an AI with 2.4 trillion parameters, boasting performance comparable to Claude Fable 5, enabling the refinement of research papers and 16 days of autonomous development - GIGAZINE

The Financial Times cited 'Claude Mythos 5,' which has garnered worldwide attention for its high ability to find vulnerabilities, noting that 'although its parameters are not publicly available, industry estimates it has 8 trillion parameters.' It then stated that ByteDance's model surpasses this.

According to sources, several Chinese AI research institutes are currently training models that rival American-made models, with ByteDance aiming to be the largest among them. The ByteDance development team is reportedly behind other companies because it is taking an independent approach rather than 'distilling' and retraining other models.

The AI development race between China and the United States is intensifying, and recently, 'Kimi K3' has been unveiled, which has achieved scores exceeding those of top-of-the-line consumer models such as GPT-5.6 Sol and Claude Fable 5 in some benchmark tests.

'Kimi K3,' a Chinese-made AI model comparable to GPT-5.6 Sol, has emerged, a massive open model with 2.8 trillion parameters - GIGAZINE