ByteDance has begun training an artificial intelligence model with up to 10 trillion parameters, positioning the Chinese tech giant to challenge the cutting-edge Mythos system from Anthropic. The move signals a major escalation in the global AI arms race as Chinese companies accelerate efforts to match US leadership.

What You Need to Know

ByteDance is in the early stages of pre-training a massive AI model that could rival the size of Anthropic's most advanced system. The model, with as many as 10 trillion parameters, would be three times larger than the biggest Chinese model released so far. This development reflects a broader trend of Chinese companies narrowing the technological gap with top US AI labs, with significant implications for global competition and innovation.

Training at Unprecedented Scale

The ByteDance model is undergoing pre-training, a phase that typically lasts three to six months, according to people familiar with the matter. After that, the system will be fine-tuned and potentially released if development stays on track. The final parameter count will be determined at a later stage, but the target approaches the scale of Anthropic's Mythos, widely considered one of the most capable AI systems in the world.

  • Parameter scale: Up to 10 trillion parameters, making it one of the largest models in development globally.
  • Training timeline: Pre-training takes three to six months before fine-tuning begins.
  • Comparative size: Three times larger than Moonshot's Kimi K3, the current largest Chinese model.

The Competitive Landscape

ByteDance's push underscores how Chinese companies are closing the gap with US leaders like Anthropic and OpenAI. Moonshot's Kimi K3, released earlier this year, was a milestone for Chinese AI, but ByteDance's effort aims to surpass it significantly. The company is leveraging its vast data resources and engineering talent to scale up rapidly, though the technical challenges of training a model this large are considerable.

The timing is critical. US export controls have restricted Chinese access to advanced chips, forcing companies like ByteDance to optimize training efficiency. Success would demonstrate that Chinese labs can compete at the frontier despite hardware constraints.

Why This Matters

If ByteDance succeeds, it would reshape the competitive dynamics of the AI industry. The ability to train a model on par with Anthropic's Mythos would give China a strong position in the race to develop general-purpose AI systems. For businesses and users, this could accelerate the availability of advanced AI tools from Chinese providers, potentially lowering costs and increasing options. However, it also raises concerns about governance, safety standards and the geopolitical implications of two superpowers fielding equally powerful AI systems. The outcome of this training effort will influence everything from corporate strategy to national policy in the years ahead.

ByteDance has not publicly commented on the effort. The model's release, if it proceeds, would mark a defining moment in the global AI landscape.