ByteDance is developing an artificial intelligence model that could be close to Anthropic's most advanced large language model yet, according to a Financial Times report on Thursday, citing insiders.
The Chinese parent of short video platform TikTok is training a model in the early stages, with a size of up to 10 trillion parameters. The model was pre-trained, which usually takes three to six months, before being fine-tuned and released, the report said.
The model's size is three times bigger than Moonshot's Kimi K3, insiders told the media outlet.
The new model could surpass Anthropic's Mythos 5, which is estimated to have about 8 trillion parameters. Meanwhile, Anthropic's Fable 5 is estimated to have about 5 trillion parameters, according to the media outlet.
ByteDance did not respond to' request for comment.
(Market Chatter news is derived from conversations with market professionals globally. This information is believed to be from reliable sources but may include rumor and speculation. Accuracy is not guaranteed.)