ByteDance Targets AI Model That Could Rival Anthropic’s Mythos in Scale
BEIJING — ByteDance is reportedly training a massive artificial intelligence model that could reach as many as 10 trillion parameters, potentially putting it close to the scale of Anthropic’s most advanced Mythos system, according to the Financial Times.
The project highlights the intensifying competition between Chinese and U.S. technology companies to build increasingly powerful frontier AI systems.
A 10 Trillion-Parameter Ambition
According to the FT report, ByteDance’s model could contain up to 10 trillion parameters.
For comparison, Reuters reports that Chinese AI startup Moonshot AI’s Kimi K3 has about 2.8 trillion parameters, meaning ByteDance’s proposed model could be more than three times larger by parameter count.
However, parameter count is only one measure of an AI system’s scale and does not necessarily determine its capabilities.
Training Is Still Underway
The ByteDance model is reportedly still in the pre-training stage, which typically takes several months before additional fine-tuning and potential release.
Reuters reported that the training process could take roughly three to six months, based on the FT’s reporting.
That means the reported 10-trillion-parameter target does not necessarily mean the model is close to being publicly available.
Challenging Anthropic’s Mythos
The reported scale would put ByteDance’s project in the same broad range as Anthropic’s Mythos.
Because Anthropic does not publicly disclose the parameter counts of its frontier models, comparisons rely on industry estimates. Reuters reported that industry estimates cited by the FT put Mythos 5 at around 8 trillion parameters and Fable 5 at around 5 trillion.
China’s AI Race Accelerates
The development comes as Chinese technology companies continue pushing toward increasingly powerful AI models while competing with leading U.S. labs.
ByteDance’s AI operation, known as Seed, has been working to develop advanced models as the global AI race increasingly focuses on computing scale, model efficiency and reasoning capabilities.
Size Doesn’t Equal Intelligence
Despite the enormous numbers involved, a larger parameter count does not automatically mean a model will outperform smaller competitors.
Model architecture, training data, compute efficiency, reasoning techniques and post-training all play major roles in determining real-world performance.
Why It Matters
If ByteDance ultimately succeeds in developing a model at the reported scale, it would represent a major milestone for China’s AI industry and further intensify competition with leading U.S. AI companies.
It would also demonstrate the extraordinary computing resources now being devoted to frontier AI development worldwide.
Conclusion
ByteDance is reportedly training an AI model targeting up to 10 trillion parameters, potentially bringing it close to the estimated scale of Anthropic’s Mythos. The model remains in pre-training, and its eventual capabilities, release timeline and final size remain uncertain.
Tags: ByteDance, AI, Artificial Intelligence, Anthropic, Mythos, China, Technology, Business News, BusinessGeco, ByteDance, AI, artificial intelligence, Mythos, Anthropic, 10 trillion parameters, China AI, Seed, frontier AI