Tencent Launches New Generation Large Model “Hunyuan Turbo” with Significant Performance Improvements

Tencent has announced the launch of its new generation large model, “Hunyuan Turbo”, which boasts significant performance improvements compared to its predecessor. The model’s training efficiency has increased by 108%, inference efficiency by 100%, and decoding speed by 20%. Additionally, the inference cost has been reduced by 50%.

Qiu Yuepeng, Vice President of Tencent, COO of the Cloud and Smart Industries Group, and President of Tencent Cloud, announced the launch of Hunyuan Turbo at the Tencent Global Digital Ecosystem Conference. The model is now available on Tencent Cloud, with input and output prices reduced by half compared to the previous generation model.

Tencent Hunyuan provides model services of various sizes on Tencent Cloud, which are fully open to enterprises and individual developers through access and use methods such as API, exclusive models, and fine-tuning models. The model offers multiple versions, including Turbo, Pro, Standard, and Lite, and provides code generation, role-playing, and Functioncall capabilities.

Tencent Hunyuan has been at the forefront of adopting the MoE structure, and its self-developed trillion-level inter-layer heterogeneous MoE structure has enabled significant improvements in both effect and performance. The model has achieved outstanding results in language understanding, text creation, mathematics, and code, with complex mathematics solving ability increased by 38% and code ability increased by 32% compared to the previous generation model.

According to the SuperCLUE benchmark for Chinese large models, Tencent Hunyuan Turbo ranked first among domestic large models in total score, thanks to its outstanding performance in multiple core tasks. The model also performed well in hard tasks centered around complex tasks and high-level reasoning, scoring 74.33 points and making it the only large model in China with a score of over 70.

Tencent Hunyuan has accumulated independent technologies from underlying computing power to machine learning platforms to upper-level applications through continuous iteration and practice. The model’s industry-leading technical strength has been recognized by many parties, including the China Electronics Society, which awarded Tencent Hunyuan’s “Key Technologies and Applications of Angel Machine Learning Platform for Large-Scale Data” the first prize for scientific and technological progress in 2023.

Tencent Hunyuan Big Model is actively promoting the implementation of internal applications to create more value for the big model. Currently, nearly 700 businesses and scenarios within Tencent have been connected, including Tencent Yuanbao, Tencent Cloud, QQ, WeChat Reading, Tencent News, and Tencent Customer Service.