ByteDance is in the early stages of training a very large artificial intelligence model that could reach roughly 10 trillion parameters, according to a Financial Times report cited by TechFlowPost. The report, which attributed the information to three people familiar with the matter, said the Chinese technology company is working on a system that may be close in scale to Anthropic’s most advanced Mythos lineup.
One of those sources said the model is now in pre-training, a phase that typically lasts three to six months. If the process goes smoothly, ByteDance would move on to fine-tuning and then release the model. The final size of the system has not yet been fixed and will be determined later in development.
The report also compared the project with existing large models. At 10 trillion parameters, the ByteDance model would be about three times the size of Kimi K3 from Moonshot AI, described in the report as the largest model released in China so far. Anthropic has not disclosed parameter counts for its own models, though industry estimates cited in the report put Mythos 5 at around 8 trillion parameters and Fable 5 at around 5 trillion.
ByteDance is training an AI model that could approach the scale of Anthropic’s most advanced Mythos system, the Financial Times reported, as Chinese companies continue to narrow the gap with leading U.S. artificial intelligence labs.
Citing three people familiar with the matter, the report said the Chinese technology group is in the early stages of training an ultra-large model that may reach 10 trillion parameters. That would make it about three times the size of Kimi K3, a model from Moonshot AI that the report described as the largest released in China so far.
One of the sources said the ByteDance model is now in pre-training, a stage that usually takes three to six months. If development proceeds smoothly, the company would then move to fine-tuning and release the model. The report added that the model’s exact size will be determined at a later stage.
Anthropic has not publicly disclosed parameter counts for its models. Industry estimates cited in the report put its most advanced Mythos 5 model at about 8 trillion parameters, while Fable 5 is estimated at about 5 trillion.
The report noted that parameter count sets a model’s basic capacity to store information, or its memory ceiling, but actual performance also depends on other factors, including data quality and training methods.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
