ByteDance Training 10-Trillion Parameter Model to Rival Anthropic

Ars Technica AI · rss · 2026-08-07

ByteDance is training a massive AI model with as many as 10 trillion parameters, aiming to narrow the gap between Chinese companies and top US labs, directly rivaling Anthropic's cutting-edge systems.

According to sources, the model's parameter count is three times larger than Moonshot’s Kimi K3, the biggest Chinese model to date. The project is currently in the pre-training stage (typically taking 3-6 months), and the exact size at release will only be determined at a later stage.

Original post →

More from Companies & People

Companies & People channel →