Zhipu AI, a unicorn in the large model field, announces the completion of financing worth several billion yuan to support technological innovation and ecosystem development. After this round of financing, Zhipu's valuation reaches 20 billion yuan, with Zhongguancun Science City Company as the lead investor, marking its third financing round in 2024. Zhipu's shareholder lineup includes well-known investment institutions such as Zhongguancun Science City, Hillhouse Capital, Qiming Venture Partners, and Junlian Capital, as well as internet giants Meituan, Alibaba, Tencent, and Xiaomi.

Recently, Zhipu AI launched its foundational large model GLM-4-Plus, showcasing visual capabilities that rival OpenAI's GPT-4, and announced that it would be available for use starting August 30. This model has made significant advancements in language foundation, text-to-image, image/video understanding, and video generation, becoming the first general video understanding model API in the country, contributing over 20 million downloads to the open-source community. GLM-4-Plus excels in language parsing, instruction execution, and long text processing.

The Zhipu AI open platform has announced the free release of the GLM-4-Flash model, aimed at promoting innovation in AI applications. This model is known for its speed and cost-effectiveness, suitable for quick response to simple tasks, supporting multi-turn dialogue, web browsing, function calls, and long-text reasoning, covering 26 languages. Through optimization techniques such as adaptive weight quantization, parallelization, batching, and speculative sampling, GLM-4-Flash significantly reduces inference costs and has acquired over 10TB of high-quality multilingual data during the pre-training stage.

Beijing Zhipu Huazhang Technology Co., Ltd. recently announced that it will provide the API interface for its GLM-4-Flash large language model free to the public to promote the popularization and application of large model technology. The GLM-4-Flash model shows significant advantages in speed and performance, particularly in inference speed, by employing adaptive weight quantization and parallel processing technologies.

In 2023, the market size for large model platforms and related applications in China reached 1.765 billion RMB, with the market still in an early stage of investment and observation. Baidu Intelligent Cloud ranked first with a market share of 19.9%, followed closely by SenseTime, while ZhiPu AI stood out as a leading startup in third place. Lu Yanxia, IDC's Research Director in China, pointed out that the market will undergo several rounds of dramatic changes in the next 2-3 years, and continuous investment in foundational large model research and enhancing base model capabilities is key for companies to maintain their competitiveness. Emerging fields such as multimodal large models and scientific computing are also on the rise.

Zhipu AI announces that the LLM GLM-4-Long, which supports ultra-long context lengths, is now available on the open platform bigmodel.cn. This model is designed specifically for processing long texts and can read the equivalent of two copies of 'Dream of the Red Chamber' or 125 papers in one go. It is widely applied in scenarios such as translating long documents, global analysis of financial reports, extracting key information, and building chatbots with ultra-long memory.

Zhipu AI announces the open source of its video generation model CogVideoX, aiming to promote the rapid development and application of video generation technology. The CogVideoX model is based on advanced large model technology and is capable of meeting commercial-grade application requirements. The currently open sourced CogVideoX-2B version requires only 18GB of GPU memory for inference at FP-16 precision and 40GB for fine-tuning, allowing inference with a single 4090 GPU and fine-tuning with a single A6000 GPU.

During its OpenDay event, Zhipu AI introduced an AI large model named Qingying, capable of generating videos from any text. Qingying can produce high-definition videos at a resolution of 1440x960 within 30 seconds. Users need only to input a text (i.e., Prompt) and choose a desired style, such as cartoon 3D, black and white, oil painting, cinematic, etc., to generate the video. Qingying is now available on the Qingyan App, accessible to all users.

Intelligent Spectrum AI has announced the open-source upgrade of the CogVLM2-Video model, a significant advancement in the field of video understanding. The CogVLM2-Video model addresses the limitations of existing video understanding models in handling missing temporal information by introducing multi-frame video images and timestamps as input to the encoder. Utilizing an automated method for constructing time localization data, it has generated 30,000 temporal-related video question and answer

Chipsight AI has announced that the GLM-4-9B and CodeGeeX4-ALL-9B models are now supported for deployment via Ollama. The GLM-4-9B is an open-source pre-trained model released by Chipsight AI, which belongs to the GLM-4 series. It demonstrates outstanding capabilities in semantics, mathematics, reasoning, coding, and knowledge in multiple aspects. The CodeGeeX4-ALL-9B is a multilingual code generation model trained based on the GLM-4-9B, which further enhances the code generation capabilities.Ol

The domestic artificial intelligence sector is witnessing new fluctuations as Zhipu AI announces a special migration plan for OpenAI API users, offering a massive token giveaway and comprehensive technical support to encourage developers to transition to its platform. This initiative not only showcases Zhipu AI's ambitions in the domestic AI market but also signifies potential changes in the global AI competitive landscape.

Zhipu AI has open-sourced CogAgent, a visual language model with 18 billion parameters. CogAgent excels in GUI understanding and navigation, achieving state-of-the-art general performance across multiple benchmark tests. The model supports high-resolution visual input and dialog Q&A, and can answer questions based on any GUI screenshot. CogAgent also supports OCR-related tasks, with its capabilities significantly enhanced through pre-training and fine-tuning.

Zhipu AI stands out in the large model race due to its advantages in practical experience, complete talent pool, and ample funding. By using GLM path training, Zhipu AI achieves higher efficiency and can understand more complex scenarios. The company attracts the attention of capital and internet giants through an open and integrated business model. Zhipu AI accelerates the implementation of large models and promotes the development of the large model ecosystem. A well-developed large model ecosystem will enhance model quality, expand application areas, and achieve customization.

["Zhipu AI has released an evaluation benchmark AlignBench for Chinese large models", "AlignBench can evaluate the alignment level of models and human intentions in multiple dimensions", "The dataset is divided into 8 major categories, including knowledge Q&A, writing generation, role-playing, and various types of questions", "Developers can use AlignBench for evaluation and score using a robust scoring model", "By logging into the AlignBench website, users can submit their evaluations."]

Zhipu AI has launched the high-quality, low-cost scoring model CritiqueLLM. Traditional evaluation metrics such as BLEU and ROUGE lack an understanding of overall semantics. CritiqueLLM proposes an interpretable and scalable text quality evaluation model that outperforms other models across 8 common tasks. CritiqueLLM generates scores through methods such as user query augmentation, reference text evaluation data collection, non-reference text evaluation data rewriting, and model training.

Zhipu AI is reported to be seeking a new round of financing with a valuation of 20 billion RMB, making it one of the highest-valued companies in the domestic large model sector. It has launched the third-generation base model ChatGLM3, which competes with GPT-4V, showcasing significantly upgraded performance and impressive rankings. Investors include major companies such as Alibaba, Meituan, Tencent, and Xiaomi, as it enters the later stages of Series B financing, aiming to raise over 2.5 billion RMB. ChatGLM3 enhances performance and reasoning capabilities, introducing a new code interpreter feature that has garnered widespread attention. Zhipu AI demonstrates its high valuation and continuous financing success.

Zhipu AI unveiled its third generation foundation model ChatGLM3 at the 2023 China Computer Conference. ChatGLM3 features new capabilities including multimodal understanding, code generation and execution, and enhanced web search. After deep optimization, ChatGLM3 has excelled in tests on Chinese and English public datasets. Zhipu Qingyan becomes the first large model product in China with code interaction capability. Zhipu AI looks forward to domestic native large models and domestic chips standing out on the international stage.

Zhipu AI has accumulated over 2.5 billion RMB in financing, with a valuation reaching 15 billion RMB, making it a rising unicorn in the large model sector. The investment landscape of Zhipu AI is essentially a network of investors who value its Tsinghua background, which brings talent and technological advantages. The participation of leading tech companies like Alibaba and Tencent in Zhipu AI reflects its leading position in the large model field. Several other companies linked to Tsinghua, such as Baichuan Intelligence, have also secured substantial financing, showcasing the influence of the Tsinghua network in the wave of large model entrepreneurship. The capabilities of foundational models will determine the future of the competition among numerous models, with Tsinghua-affiliated companies in a favorable position.

Baichuan Intelligence and Zhipu AI respectively secured hundreds of millions of dollars and billions of RMB in financing. Zhipu AI is regarded as a seed player of OpenAI in China, applying large model technology to industrial scenarios. The financing of Zhipu AI has received support from several well-known institutions and investors.