ByteDance, the Chinese technology group that owns TikTok, is reportedly training an artificial intelligence model with as many as 10 trillion parameters, Free Press Journal reported on August 07, citing the FT and people familiar with the matter. The model could approach the scale of Anthropic's Mythos system, but the report said the details could not be independently verified and ByteDance had not immediately responded to a request for comment.

A parameter count means the numerical settings a model learns from training data to recognise patterns, generate responses and perform tasks; it is often used as a rough measure of model scale, though not a direct measure of capability. At 10 trillion parameters, ByteDance's model would be more than three times the size of Moonshot AI's Kimi K3, which has 2.8 trillion parameters and is among the largest models released by a Chinese AI lab so far.

The FT cited industry estimates that put Anthropic's Mythos 5 at around 8 trillion parameters and Fable 5, its more widely available guardrailed sibling, at about 5 trillion parameters. Anthropic does not officially disclose parameter counts for Mythos, Fable or any of its models, so those figures remain industry estimates; the article said the same caveat applies to parameter counts attributed to other US models such as OpenAI's GPT-5.5.

ByteDance's model is in pre-training (initial learning from large datasets), a stage that typically takes three to six months, according to the report. The model would then move to fine-tuning before any possible public release, and the final parameter count has not been fixed and could change before training ends.

The development follows reports that ByteDance founder Zhang Yiming told employees to avoid relying on AI distillation techniques purely for short-term gains. That message came after a senior US official alleged that Moonshot AI had distilled Anthropic's Fable model while building Kimi K3.

Mythos is Anthropic's frontier model class above its Opus tier and is described as drawing attention for agentic coding, reasoning and cybersecurity capabilities. Access to Mythos has largely been limited to trusted partners under Anthropic's Project Glasswing because of misuse concerns in hacking and vulnerability exploitation, while Fable 5 is the safety-hardened version made more broadly available.

ByteDance, the Chinese technology group that owns TikTok, is reportedly training an artificial intelligence model with as many as 10 trillion parameters, Free Press Journal reported on August 07, citing the FT and people familiar with the matter. The model could approach the scale of Anthropic's Mythos system, but the report said the details could not be independently verified and ByteDance had not immediately responded to a request for comment.

A parameter count means the numerical settings a model learns from training data to recognise patterns, generate responses and perform tasks; it is often used as a rough measure of model scale, though not a direct measure of capability. At 10 trillion parameters, ByteDance's model would be more than three times the size of Moonshot AI's Kimi K3, which has 2.8 trillion parameters and is among the largest models released by a Chinese AI lab so far.

The FT cited industry estimates that put Anthropic's Mythos 5 at around 8 trillion parameters and Fable 5, its more widely available guardrailed sibling, at about 5 trillion parameters. Anthropic does not officially disclose parameter counts for Mythos, Fable or any of its models, so those figures remain industry estimates; the article said the same caveat applies to parameter counts attributed to other US models such as OpenAI's GPT-5.5.

ByteDance's model is in pre-training (initial learning from large datasets), a stage that typically takes three to six months, according to the report. The model would then move to fine-tuning before any possible public release, and the final parameter count has not been fixed and could change before training ends.

The development follows reports that ByteDance founder Zhang Yiming told employees to avoid relying on AI distillation techniques purely for short-term gains. That message came after a senior US official alleged that Moonshot AI had distilled Anthropic's Fable model while building Kimi K3.

Mythos is Anthropic's frontier model class above its Opus tier and is described as drawing attention for agentic coding, reasoning and cybersecurity capabilities. Access to Mythos has largely been limited to trusted partners under Anthropic's Project Glasswing because of misuse concerns in hacking and vulnerability exploitation, while Fable 5 is the safety-hardened version made more broadly available.