THE CLAIM. The Financial Times reported, citing three people, that ByteDance is pretraining a model with up to 10 trillion parameters to rival Anthropic's Mythos.

THE CHECK. the report holds. Multiple outlets corroborate it. But 10 trillion is described as an upper bound still under consideration, not a final spec, and there are zero published evals. The 'rivals Mythos' line compares it to an unofficial 8 trillion estimate that Anthropic has never confirmed, because Anthropic does not publish parameter counts at all.

THE TWIST. parameter count is not capability. Kimi K3 lists 2.8 trillion parameters but activates only 104 billion per token, and nobody has said whether ByteDance's number is dense or sparse. As one write-up put it, ten trillion is the upper bound of scale under consideration, not a finished performance benchmark. Bigger is bigger. It is not automatically better.