Chinese startup Z.ai launches powerful open source GLM-4.5 model family with Power Point creation

The summer of 2025 ushers in exciting developments in the world of AI, as the Chinese startup Z.ai introduces its latest offerings: the GLM-4.5 and GLM-4.5-Air models. These new open-source large language models (LLMs) are designed to excel in AI reasoning, coding, and agentic behavior, positioning themselves as formidable contenders against leading proprietary models in the U.S. According to Z.ai’s blog, the flagship GLM-4.5 model not only matches but often outperforms several top-tier models, including Claude 4 Sonnet and Gemini 2.5 Pro. It has secured an impressive third place across a range of competitive evaluations, while its lighter counterpart, the GLM-4.5-Air, ranks among the top six performers in its category. Both models boast dual operational modes: one tailored for complex reasoning and tool usage, and another for quick-response scenarios. One of the standout features of these models is their ability to automatically generate complete PowerPoint presentations from a simple title or prompt. This capability makes them particularly useful for various applications, including education, meeting preparation, and internal reporting. Additionally, they can assist in creative writing, emotionally aware copywriting, and script generation, aiding in the development of branded content for social media. The models also support virtual character development and dialogue systems, making them ideal for customer support, role-playing, and storytelling. While both GLM-4.5 and GLM-4.5-Air are designed for reasoning and coding tasks, the latter is optimized for teams looking for a lighter and more cost-effective solution without compromising speed or resource demands. Z.ai has also included several specialized models within the GLM-4.5 family, such as GLM-4.5-X and GLM-4.5-Air X, which focus on ultra-fast inference. The GLM-4.5-Flash variant is free and optimized specifically for coding and reasoning tasks. These models are readily accessible via Z.ai’s API, as well as through platforms like Hugging Face and Model Scope, providing developers with flexible integration options. The GLM-4.5 models are released under the Apache 2.0 license, which encourages developers to use, modify, and redistribute the models for research and commercial purposes. Z.ai has set competitive pricing for its cloud-based API, charging approximately $0.11 to $0.28 per million tokens for input and output, depending on the scale of usage. The GLM-4.5 has earned accolades for its performance, ranking third in agentic, reasoning, and coding evaluations, trailing only behind OpenAI’s GPT-4 and X AI’s Grok 4. With accuracy rates of 98.2% on MATH 500 and competitive results in various coding benchmarks, it demonstrates a strong capability in AI tasks. This release comes amid a broader trend of competitive open-source AI model launches in China, with notable contributions from other companies like Alibaba. As Chinese startups like Z.ai and Alibaba continue to innovate and release powerful models, they are positioning themselves as significant players in the global AI landscape. Founded in 2019, Z.ai has rapidly gained prominence in the AI sector, securing over $1.5 billion in funding from major investors, including Alibaba and Tencent. The launch of GLM-4.5 coincides with the World Artificial Intelligence Conference in Shanghai, where the company showcased its advancements. For organizations looking to harness the power of AI, the GLM-4.5 family offers a high-performing, open-source foundation model that can be tailored and scaled according to specific needs, without being constrained by proprietary systems. This flexibility makes it an attractive option for teams aiming to drive innovation while managing operational costs.