Key takeaways
- Tencent launches Hunyuan Turbo S, outperforming competitors in several AI tasks.
- Model features hybrid-mamba-transformer architecture for enhanced efficiency and reduced costs.
- Achieves 44% reduction in first-word delay, mimicking human intuition in responses.
- Pricing set at 0.8 yuan per million tokens for input and 2 yuan for output.
- Claims first lossless application of Mamba architecture in a super-large Mixture of Experts model.
Chinese tech giant Tencent has released its new AI model, Hunyuan Turbo S, which it says can answer queries faster than the DeepSeek-R1 model. The model is available on the official Tencent Cloud website and can be accessed via API.
The Hunyuan Turbo S doubles the output speed and reduces the first-word delay by 44%, the company announced on its official WeChat channel.
Tencent said that the fast-thinking model is analogous to human intuition, which often results in rapid responses compared to rational thinking.
However, the company said Hunyuan Turbo S efficiently solves problems by fusing long and short thinking chains.
The model uses an innovative hybrid-mamba-transformer fusion architecture. It optimises efficiency by lowering the computational complexity of the conventional transformer, minimising KV-Cache storage usage, and reducing training and inference costs.
The company also said that the model leverages Mamba's efficiency in processing long sequences while preserving the Transformer's strength in capturing complex contextual relationships.
Tencent also claims this is the first time the Mamba architecture has been applied losslessly to a super-large Mixture of Experts (MoE) model.
Tencent also released benchmark results, and the model is better, if not on par with other large language models like DeepSeek-V3, Claude 3.5 Sonnet, and GPT-4o—in mathematics, coding, and reasoning tasks.
The Hunyuan Turbo S's input API price is 0.8 yuan ($0.11) per million tokens, and its output price is 2 yuan ($0.28) per million tokens.
Source: Tencent
Amid the rise of DeepSeek, the competition in China’s AI ecosystem is heating up. Recently, Alibaba introduced a preview of the Qwen QwQ-Max reasoning model and committed to a $52 billion investment in AI infrastructure over the next three years.
It was also reported that DeepSeek plans to release its next reasoning model, the DeepSeek R2, ‘as early as possible’. The company initially planned to release it in early May but is now considering an earlier timeline.
The model is expected to produce ‘better coding’ and reason in languages beyond English.
Note: The headline has been updated to provide better clarity.
