NewsBite analysis · AI, Business
The Decoder
China's Largest AI Model Is Being Developed at Bytedance
Bytedance is currently training an AI model that is projected to have ten trillion parameters, placing it far ahead of other domestic models like Moonshot's Kimi K3. This advancement means that Bytedance is not only solidifying its standing in the Chinese AI market but is also positioning itself alongside major international players such as Anthropic. The pretraining phase, which typically spans three to six months, has reportedly avoided methods that rely on outputs from existing companies, indicating a focus on building a unique model. The ambition is clear from Bytedance's founder, who has urged the team to pursue world-leading capabilities.
MonitorConfidence: HighShort term horizonData governance risk: High
Published 7 Aug 2026, 14:10 UTC1 min read
Visit original article ↗Lead image — The Decoder
Summary
TL;DR
- Bytedance's AI model aims for ten trillion parameters, the largest in China.
- This model exceeds the capabilities of existing Chinese rivals.
- Bytedance's approach emphasizes quality data and unique training methods.
- It places Bytedance in direct competition with global AI leaders.
- The pretraining phase is on track for completion within months.
Interesting facts
- The model's size aims to rival Anthropic's Mythos 5 with eight trillion parameters.
- Bytedance has avoided using distillation techniques for over a year.