Tencent Launches Upgraded T1 Reasoning Model in China's AI Arms Race

Chinese tech giant Tencent officially launched the upgraded version of its T1 reasoning model, intensifying competition in China's burgeoning AI sector. The company's announcement, posted on its official WeChat account, highlights the model's enhanced capabilities and faster response times.

"T1 can 'keep the content logic clear and the text neat and clean', while the hallucination rate is 'extremely low'," the post stated, emphasizing the model's improved accuracy and clarity.

The launch follows the emergence of DeepSeek's AI models, which have garnered attention for offering comparable or superior performance to Western systems at significantly lower costs. This has spurred increased competition within China's AI landscape.

Tencent's T1 model was previously released in a preview version through platforms such as its AI assistant application, Yuanbao. The official release will leverage the company's Turbo S foundational language model, unveiled in late March, which Tencent claims boasts faster processing speeds than DeepSeek's R1 model.

A chart accompanying the announcement compares T1 to DeepSeek R1, showing Tencent's model outperforming on certain knowledge and reasoning benchmarks.

Tencent has been actively increasing its investments in AI in recent months. On Thursday, the company announced plans to further increase capital expenditure in 2025, following substantial AI spending throughout 2024.