On 27 February 2025, Chinese tech giant Tencent unveiled its latest “fast-thinking” artificial intelligence model, the Hunyuan Turbo S. Compared to the DeepSeek R1 model, it boasts a 44% reduction in response time and twice the word generation speed. The new model adopts an innovative Hybrid-Mamba-Transformer architecture, which significantly reduces computational complexity and is capable of handling both intuitive rapid responses and tasks requiring complex problem-solving with high efficiency.
The performance of Hunyuan Turbo S in independent tests is on par with leading industry models such as DeepSeek V3, GPT-4o, and Claude in areas including knowledge, mathematics, and creative content generation. The model is also cost-effective for developers: Tencent charges 0.8 yuan (approximately 11 cents) per million input tokens and 2 yuan (around 27 cents) per million output tokens. According to Tencent’s announcement, Hunyuan Turbo S will serve as the foundation of the company’s future AI model family, which will include specialised models for reasoning, long-form text processing, and code generation.
A fierce race has emerged in the field of Chinese AI development, beginning with the launch of the DeepSeek-V3 model in December 2024, which matched the performance of leading Western models at a fraction of the cost. This was followed by Alibaba’s updated Qwen2.5-Max model, which outpaced both Meta’s Llama and DeepSeek-V3 in terms of speed. As the next step, Hunyuan Turbo S introduces a unique “fast and slow thinking” approach, combining rapid response capabilities (fast thinking) with deeper analytical reasoning (slow thinking). It achieved outstanding results in benchmarks such as MMLU, Chinese-SimpleQA, and C-Eval, while also demonstrating exceptional mathematical performance in the MATH and AIME2024 tests.
Sources:
1.
Syed Sadain Gardazi
2.
Tencent
3.
Mehul Gupta
4.
Clyde Morgan
