ByteDance bets 10 trillion parameters on catching Anthropic — because size is everything, right?

ByteDance is reportedly training a super-large AI model with an estimated 10 trillion parameters, aiming to compete with Anthropic's advanced Mythos system. The model is in early pre-training, a phase expected to take three to six months before fine-tuning and release. For context, Anthropic's Mythos 5 is estimated at around 8 trillion parameters, while China's current largest released model, Moonshot AI's Kimi K3, is roughly a third of ByteDance's target. The final parameter count has not been finalized, and actual performance will depend on data quality and training methodology, not just scale.