Zhipu recently completed a new round of funding, raising approximately $410 million. New investors include several strategic partners and state-owned capital, with existing investors like Legend Capital continuing to participate. This marks the second known round of funding for Zhipu in recent times.
Reportedly, the funds will be used for the development of the Zhipu GLM model series. The goal is to expand from simple question answering to tackling more complex reasoning and multimodal tasks, enhancing the development of the industry ecosystem.
In 2024, Zhipu’s business continued to grow rapidly, with commercial revenue more than doubling compared to 2023. Currently, Zhipu’s valuation has exceeded $20 billion.
In terms of product business, Zhipu’s all-modal AI assistant app “Zhipu Qingyan” has over 25 million users and generates annual revenue of approximately $1.37 million. Their MaaS open platform bigmodel.cn is used by 700,000 enterprises and developers, with daily token consumption increasing 150 times and API annual revenue growing more than 30 times year-over-year.
Zhipu consistently leads in large model innovation. Reportedly, Zhipu’s open-source models like ChatGLM have been downloaded over 30 million times globally, earning more than 150,000 GitHub stars. This makes Zhipu the company with the most stars in China’s large model field.
Zhipu is the only Chinese company selected for TechCrunch’s Unicorn Board, listing the top 15 new AI unicorns globally. The company was also named one of Forbes China’s Top 50 Innovative Companies and one of the most popular AI organizations on the Hugging Face platform.
At the end of 2020, Zhipu developed the GLM pre-training architecture. In 2021, they completed the training of the GLM-10B model with tens of billions of parameters. By 2022, they developed and open-sourced the GLM-130B, a massive bilingual pre-training model in Chinese and English. In 2023, they launched the ChatGLM, a dialogue model with hundreds of billions of parameters, and upgraded it twice. The open-source version, ChatGLM-6B, allows developers to fine-tune and deploy the model locally.
In 2024, Zhipu continued its rapid development, releasing a series of large model products and technologies:
January: Released the next-generation base model GLM-4, with significantly improved performance, supporting longer contexts, stronger multi-modal capabilities, faster inference speed, higher concurrency, and reduced inference costs.
June: Open-sourced the GLM-4-9B model, supporting long text processing and multiple languages, and also open-sourced the GLM-4V-9B visual model, with multi-modal capabilities comparable to GPT-4V.
July: Launched the video generation model “Qingying,” offering text and image to video services, capable of generating a 6-second video in 30 seconds, accurately replicating physical world movements.
August: Released China’s first real-time inference large model across text, audio, and video modalities, GLM-4-Videocall, enabling real-time video calls between AI and humans. Introduced the next-generation base model GLM-4-Plus, with comprehensive performance improvements, comparable to the international benchmark GPT-4o.
October: Released the GLM-4-Voice end-to-end emotional voice model, integrated into the Qingyan app, supporting emotional understanding, expression, multiple languages, and dialects. Launched the beta version of AutoGLM, capable of simulating human smartphone operations to perform various tasks beyond API calls.
November: Released an upgraded version of AutoGLM, capable of executing operations over 50 steps long, performing tasks across apps, and supporting browser-based website automation. Launched the GLM-PC beta, based on the multi-modal model CogAgent, exploring “driverless” PC technology to perform various computer operations for users. Upgraded the video model CogVideoX, supporting 10-second, 4K, 60-frame ultra-high-definition video generation, with improved human motion and physical world simulation, and open-sourced CogVideoX v1.5-5B and CogVideoX v1.5-5B-I2V models.
Zhipu’s business scope covers multiple industries, including smart cars, manufacturing, consumer goods, finance, government services, healthcare, gaming, and cultural tourism. The recently launched Agent further advances the application of large models in smart devices.
Zhipu’s globalization strategy is also accelerating, with regional headquarters, subsidiaries, and research centers established in countries and regions such as the United States, the United Kingdom, and France. It has signed a frontier AI safety commitment with world-class large model companies like OpenAI, Google, Microsoft, and Meta.
The vision of Zhipu is “to make machines think like humans,” dedicated to creating a new generation of cognitive intelligence large models.
Source from ifanr
Disclaimer: The information set forth above is provided by ifanr.com, independently of Chovm.com. Chovm.com makes no representation and warranties as to the quality and reliability of the seller and products. Chovm.com expressly disclaims any liability for breaches pertaining to the copyright of content.
