Skip to content
LargeLanguageModel.com.tr
All articles

Zhipu Unveils Groundbreaking Open-Source AI Model: GLM-4.5 Released

1 min read

Chinese AI startup Zhipu has launched its next-generation open-source language model, GLM-4.5. The flagship version comes with 355 billion parameters, while the lighter GLM-4.5-Air offers similar capabilities with only 12 billion active and 106 billion total parameters, making it more accessible for users with limited hardware. Designed for intelligent agent applications, GLM-4.5 is now available to developers worldwide.

Zhipu, one of China’s leading artificial intelligence startups, has officially launched its advanced open-source language model, GLM-4.5. The launch follows the company’s recent rebranding to Z.ai in English, and is expected to be a milestone in the development of intelligent agent applications.


GLM-4.5 is available in two versions. The flagship model boasts an impressive 355 billion parameters, promising high-performance data processing at scale. Alongside, the GLM-4.5-Air variant is offered for users with limited hardware resources. While it has a total of 106 billion parameters, only 12 billion are active, enabling similar capabilities with a dramatically reduced computational footprint.


According to Zhipu, GLM-4.5 is optimized for intelligent agent technologies and is released as open source for global developers. Users can access API services for the model via the Z.ai API Platform (global) or the Zhipu AI Open Platform (mainland China).


This move by Zhipu signals China’s growing competitiveness in the global AI race and highlights the country’s commitment to strengthening the open-source ecosystem.

Share
Artificial Intelligence

Alibaba’s Open-Source Move: Wan 2.2 Brings Cinematic AI to All

Alibaba’s AI platform, Wan, has launched its latest model, Wan 2.2, designed to eliminate creative barriers using artificial intelligence. With support for both text-to-video and image-to-video generation at 720P and 24fps, the open-source platform stands out for its user-friendly approach. Wan 2.2 enables the rapid production of near-cinematic quality videos even on consumer-grade graphics cards.

1 min
Artificial Intelligence

Google DeepMind unveils “Genie 3,” a real-time interactive world model

Google DeepMind has unveiled Genie 3, a new world model that generates navigable, interactive environments from text prompts at 720p resolution and 24 fps, maintaining consistency for minutes. The model offers real-time control across natural physics, animation, and historical settings, and introduces “promptable world events” to alter scenes via text, such as changing weather or adding objects. Released as a limited research preview, Genie 3 advances long-horizon environmental consistency and action-conditioned simulation, marking a significant step toward AGI.

2 min