news

The latest evaluation of the Chinese large model is released: Hunyuan is the first

2024-08-05

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

According to Fast Technology on August 5, in the latest released Chinese multimodal large model SuperCLUE-V benchmark evaluation, Tencent Hunyuan Large Model ranked first in China and remained firmly in the excellent leader quadrant.

This evaluation focuses on the key ability of large models to understand the complex real world, namely multimodal understanding, commonly known as "text from pictures".

Multimodal understanding requires models to be able to accurately identify image elements, understand the relationships between them, and generate natural language descriptions.

This capability not only tests the accuracy of image recognition, but also reflects a comprehensive understanding of the scene and a deep insight into details. It is an important criterion for measuring the model's understanding of the complex real world.

In this evaluation, a total of 12 representative multimodal understanding big models from home and abroad participated. Tencent's Hunyuan big model showed significant advantages in both multimodal basic capabilities and application capabilities, and obtained a high score of 71.95 in total.

The evaluation criteria cover dimensions such as accuracy of understanding, relevance of responses, and depth of reasoning, ensuring the scientificity and fairness of the evaluation.

Tencent's Hunyuan large model is closely following the top overseas models in terms of basic capabilities, with an overall score only slightly lower than GPT-4o, and outperforming CLaude3.5-Sonnet and Gemini-1.5-Pro.

In terms of application capabilities, Tencent's Hunyuan Big Model demonstrates great potential for practical application by virtue of its deep understanding of the Chinese context and its comprehensive capabilities in multiple fields such as general knowledge, common sense, and images.