Zhipu AI Unveils New Generation of Basic Models with Impressive Advancements

TapTechNews August 29th news, Zhipu AI released a new generation of basic models at the KDD 2024 conference, claiming to have reached the international first-tier level in the corresponding fields, and announced the free opening of the GLM-4-Flash API on the Maas platform:

The language model GLM-4-Plus : Its performance in language understanding, instruction following, and long text processing has been comprehensively improved.

The text-to-image model CogView-3-Plus : Has performance close to the current top models such as MJ-V6 and FLUX.

The image/video understanding model GLM-4V-Plus: It has excellent image understanding ability and video understanding ability based on time acquisition. This model will be launched on the open platform (bigmodel.cn) and become the first general video understanding ability model API in China.

The video generation model CogVideoX : After the 2B version was released and open sourced, the 5B version was also officially open sourced, and its performance was further enhanced.

GLM-4-Flash API: Inference service is free and provides fully powerful service.

Zhipu AI Unveils New Generation of Basic Models with Impressive Advancements_0

In addition, Zhipu also announced the launch of the video call function on the Qingyan App, which is the first video call service open to C-end users in China.

Zhipu AI Unveils New Generation of Basic Models with Impressive Advancements_1

Zhipu said that the GLM-4-Plus basic model has reached the same level as the first-tier models such as GPT-4o in various indicators. In addition, during the training process of the GLM-4-Plus model, Zhipu significantly enhanced the model's performance in long text reasoning by adopting a more accurate long and short text data mixing strategy.

Zhipu AI Unveils New Generation of Basic Models with Impressive Advancements_2

For CogView-3-Plus, Zhipu uses the Transformer architecture to replace the traditional UNet architecture to train the diffusion model, and has in-depth studied the noise planning of the diffusion model, and its effect is currently close to the first-class models such as MJ-V6 and FLUX.

Zhipu AI Unveils New Generation of Basic Models with Impressive Advancements_3

At present, the GLM-4-Plus model has been officially launched on the open platform (TapTechNews note: bigmodel.cnb), providing API services to the outside world; at the same time, it will also be launched on the Qingyan App soon; the CogView-3-Plus model has been officially launched on the Qingyan App for everyone to use.

Zhipu AI Unveils New Generation of Basic Models with Impressive Advancements_4

Likes