Beijing Zhipu Huazhang Technology Co., Ltd. released its new generation base model and application services on August 29, 2024, and demonstrated it in detail at the KDD2024 conference. This update covers multiple modalities such as language, images, and videos, and launches a new application for C-end users, marking that Zhipu has made significant progress in the field of artificial intelligence, and its technical strength and innovation capabilities have been further improved. . Below is a detailed explanation of this update.
At the KDD2024 conference, Zhipu released a new generation of base models including the language model GLM-4-Plus, the Vincent graph model CogView-3-Plus, the image/video understanding model GLM-4V-Plus, and the video generation model CogVideoX. . These models have reached international leading levels in their respective fields. The performance of the GLM-4-Plus model has been comprehensively improved in terms of language understanding, instruction following, and long text processing, and is on par with first-tier models such as GPT-4o. The CogView-3-Plus model uses the Transformer architecture to replace the traditional UNet architecture, which optimizes the model effect, and its performance is close to first-line models such as MJ-V6 and FLUX. The GLM-4V-Plus model has high-quality image understanding and video understanding capabilities, becoming the first domestic general video understanding model API. After the release of the 2B version, the CogVideoX model further opened up the 5B version, with enhanced performance, becoming the leader among current open source video generation models. In addition, Zhipu launched China's first video call service for C-end users on the "Qingyan APP". The service spans text, audio and video modes and has real-time reasoning capabilities, providing users with a smooth interactive experience. . Zhipu also announced the free use of GLM-4-Flash API, which has advantages in speed and performance, allowing users to build exclusive models and applications quickly and for free. At the same time, in order to meet the needs of different users, Zhipu provides model fine-tuning functions. Zhipu said it will continue to move forward, making machines think like humans and bringing more advanced technologies and services to users.
In addition, Zhipu launched China's first video call service for C-end users on the "Qingyan APP". This service spans text, audio and video modes, and has real-time reasoning capabilities, providing users with a smooth interactive experience. .
Zhipu also announced the free use of GLM-4-Flash API, which has advantages in speed and performance, allowing users to build exclusive models and applications quickly and for free. At the same time, in order to meet the needs of different users, Zhipu provides model fine-tuning functions.
Zhipu said it will continue to move forward, making machines think like humans and bringing more advanced technologies and services to users.
Major updates:
Language base model GLM-4-Plus: Its performance has been comprehensively improved in terms of language understanding, instruction following, and long text processing, maintaining the international leading level.
Vincent diagram base model CogView-3-Plus: has performance close to the current best models such as MJ-V6 and FLUX.
Image/Video Understanding Base Model GLM-4V-Plus: It has excellent image understanding capabilities and has video understanding capabilities based on time perception. The model will be launched on the open platform (bigmodel.cn) and become the first general video understanding model API in China.
Video generation base model CogVideoX: After the 2B version was released and open sourced, the 5B version was also officially open sourced. Its performance has been further enhanced and it is the best choice among the current open source video generation models.
"Qingyan APP" launched video calling: the first domestic video calling service open to C-end users. The video calling function of "Qingyan APP" spans text, audio and video modes, and has real-time reasoning capabilities.
GLM-4-Flash API: The inference service is completely free and provides fine-tuning services.
Video call service application link:
https://zhipu-ai.feishu.cn/share/base/form/shrcnqpIx9q5ILEFeT2cPNhyuSf
All in all, Zhipu Huazhang’s technological update demonstrates its strong strength and continuous innovation capabilities in the field of artificial intelligence, brings more advanced technologies and services to users, and injects new vitality into the development of the artificial intelligence industry.