Tencent Releases Next-Generation Large Model "Tencent Hybird Turbo"

In the 2024 Tencent Global Digital Ecology Conference, Tencent unveiled its latest generation model, Tencent Hybrid Turbo. This model is based on the MoE architecture design and has improved inference efficiency by 100% compared to its predecessor, while reducing inference costs by 50%.

For external users, the price of Tencent Hybrid Turbo has been reduced by 50% compared to Hybrid Pro. The output price is set at 0.05 yuan per thousand tokens, while the input price is 0.015 yuan per thousand tokens.

According to the August 2024 report released by SuperCLUE on Chinese large-scale models benchmark evaluation, Tencent Hybrid Turbo-Preview performed exceptionally well in the tests and ranked first in both arts and sciences fields. In terms of Hard task scores, Tencent Hybrid Turbo-Preview achieved a score of 74.33, becoming the only large-scale model in China to score over 70, which is similar to the score of ChatGPT-4o.

Tencent has integrated Tencent Hybrid models into nearly 700 businesses and application scenarios, including Tencent Yuanbao, Tencent Cloud, QQ, WeChat Reading, Tencent News, and Tencent Customer Service, among many other products and services. In addition, Tencent's collaboration SaaS products have also fully integrated Tencent Hybrid models.

The cloud version of Tencent Hybrid offers various options such as Turbo, Pro, Standard, and Lite. As for exclusive models, it provides various services including code generation, role-playing, and function calls.