Zhipu AI, a leading company in China's artificial intelligence field, has once again stirred up the industry. According to AIbase, Zhipu AI recently officially open-sourced its new general vision model GLM-4.1V-Thinking. This 900 million parameter multimodal reasoning model, with outstanding performance and wide application scenarios, not only set new records in multiple authoritative evaluations but also demonstrated powerful capabilities comparable to or even surpassing 7.2 billion parameter models. The following is the latest information compiled by AIbase, giving you an in-depth understanding of this breakthrough technology.
Introducing Chain-of-Thought Reasoning, Performance Significantly Improved
GLM-4.1V-Thinking is deeply optimized based on Zhipu AI's previous GLM-4V architecture, introducing an innovative Chain-of-Thought Reasoning mechanism. This mechanism significantly enhances the model's performance in complex cognitive tasks, allowing it to process multimodal inputs such as images, videos, and documents more efficiently. According to AIbase, the model achieved the highest scores of 1 billion parameter-level models in 23 out of 28 authoritative evaluations (such as MMStar, MMMU-Pro, ChartQAPro, OSWorld, etc.), and in 18 of these, it performed equally well or even surpassed the Qwen-2.5-VL72B model with a larger parameter scale, demonstrating its impressive reasoning capabilities.
Comprehensive Multimodal Capabilities Empower Diverse Industries
GLM-4.1V-Thinking supports a context length of up to 64K and an image resolution of 4K, while also being capable of handling multilingual scenarios with both Chinese and English. Whether it's long video understanding, image QA, subject problem-solving, text recognition, document interpretation, image localization (Grounding), GUI agent operations, or code generation, this model can handle them effortlessly. Its open-source nature further lowers the usage threshold, allowing it to run on a single 3090 GPU, and the free commercial license provides broad application opportunities for enterprises and developers. AIbase believes that this combination of flexibility and high performance will greatly promote the practical application of AI technology in industries such as education, finance, and healthcare.
Open-Source Strategy, Leading the Global AI Competition
Zhipu AI chose to fully open-source GLM-4.1V-Thinking and provides model weights and demos through the Hugging Face platform, demonstrating its determination to promote the popularization of AI technology. AIbase noticed that Zhipu AI has been actively involved in open-source initiatives in recent years. The GLM series models have been downloaded over 30 million times globally, becoming an important part of the Chinese AI ecosystem. The open-sourcing of GLM-4.1V-Thinking not only provides developers with a high-performance multimodal reasoning tool but also ensures commercial flexibility through the MIT license, further strengthening Zhipu AI's competitiveness in the global AI field.
Direct Competition with Top Global Models
In performance comparisons, GLM-4.1V-Thinking has demonstrated remarkable capabilities. According to AIbase's comprehensive evaluation data, the model performs exceptionally well in various complex tasks, especially in high-difficulty scenarios such as STEM subject problems and long-document understanding, where some of its performance even exceeds OpenAI's GPT-4o model. This significant advancement indicates that Zhipu AI has joined the ranks of global leaders in multimodal reasoning and is now competing directly with international giants like OpenAI and Google.
A New Chapter in the Rise of Chinese AI
As one of the "New Four Tigers" in China's AI field, Zhipu AI is reshaping the global AI landscape through continuous technological innovation and an open ecological strategy. AIbase believes that the release of GLM-4.1V-Thinking not only reflects Zhipu AI's technical strength but also marks an important voice for China's AI industry on the global stage. In the future, as more developers build innovative applications based on GLM-4.1V-Thinking, the international influence of Chinese AI will further expand.
Conclusion
Zhipu AI's GLM-4.1V-Thinking, with its powerful multimodal reasoning capabilities and open-source characteristics, brings new possibilities to the global AI community. AIbase will continue to follow Zhipu AI's latest developments and bring you more cutting-edge technology reports. Let's look forward to how this model will drive transformation across various industries!