%20--%3e%3cdefs%3e%3cstyle%3e%20.st0%20{%20fill:%20%23061b40;%20}%20.st1%20{%20fill:%20%23306af1;%20}%20.st2%20{%20fill:%20%235ce5cf;%20}%20%3c/style%3e%3c/defs%3e%3cg%3e%3cpath%20class='st0'%20d='M55,10.5h9v3h-9v9h12V7.5h-12v3ZM64,19.5h-6v-3h6v3Z'/%3e%3cpolygon%20class='st0'%20points='69%2016.5%2078%2016.5%2078%2019.5%2069%2019.5%2069%2022.5%2081%2022.5%2081%2013.5%2072%2013.5%2072%2010.5%2081%2010.5%2081%207.5%2069%207.5%2069%2016.5'/%3e%3cpolygon%20class='st0'%20points='95%2010.5%2095%207.5%2083%207.5%2083%2022.5%2095%2022.5%2095%2019.5%2086%2019.5%2086%2016.5%2095%2016.5%2095%2013.5%2086%2013.5%2086%2010.5%2095%2010.5'/%3e%3cpath%20class='st0'%20d='M40,1.5v21h11.6l1.4-1.4v-7.6h0c0,0-1.4-1.5-1.4-1.5l1.4-1.4V2.9l-1.4-1.4h-11.6ZM50,19.5h-7v-6h7v6ZM50,10.5h-7v-6h7v6Z'/%3e%3c/g%3e%3cpath%20class='st1'%20d='M23.1,24L14.7,4.8l-4.9,11.2h3.8l-1.8,4H3.3L12.1,0H2C.9,0,0,.9,0,2v20c0,1.1.9,2,2,2h21.1Z'/%3e%3cpath%20class='st2'%20d='M34,0h-16.8l10.6,24h6.2c1.1,0,2-.9,2-2V2C36,.9,35.1,0,34,0ZM32.5,20h-4V4h4v16Z'/%3e%3c/svg%3e)
Tencent has recently announced the open-source release of its new language model "**Hunyuan-A13B**," which introduces an innovative "**dynamic reasoning method**," enabling it to intelligently switch between fast and deep "thinking" based on task complexity.
Key Highlights: Dynamic Reasoning and MoE Architecture
The core advantage of "**Hunyuan-A13B**" lies in its ability to **adjust the depth of reasoning in real-time**. For simple queries, the model uses a **fast mode**, responding quickly with minimal reasoning steps; for more complex questions, it can activate a **deep reasoning process involving multiple steps of thinking**. Users can also manually control this behavior using specific commands, such as "**/think**" to enable deep mode or "**/no_think**" to disable it.
The model employs a **Mixture of Experts (MoE) architecture**, with a total parameter count of **80 billion**, but only **13 billion parameters** are active during actual inference, which helps improve efficiency. Additionally, "**Hunyuan-A13B**" supports a context window of up to **256,000 tokens**, significantly enhancing its ability to handle long texts.
Training Data and Performance
According to Tencent's technical report, "**Hunyuan-A13B**" was trained on **20 trillion tokens** and optimized for reasoning tasks and broader use cases. To enhance the reliability of the model in scientific tasks, Tencent specifically collected **250 billion tokens** from the **STEM (Science, Technology, Engineering, and Mathematics)** field for training, including data such as math textbooks, exams, GitHub open-source code, logic puzzles, and scientific texts from **high school to university levels**.
Tencent claims that its "**Hunyuan-A13B-Instruct**" version is comparable to leading models like OpenAI, Deepseek, and Alibaba Qwen. The report states that in the **2024 American Invitational Mathematics Examination (AIME)**, "**Hunyuan-A13B**" achieved an accuracy rate of **87.3%**, outperforming OpenAI o1's 74.3%.
