Zhipu AI launches the third-generation large base model ChatGLM3 to adapt to more domestic chips

The news on October 27, 2023 is that Zhipu AI released a new self-developed third-generation large base model ChatGLM3 and related series of products at the China Computer Conference (CNCC). This release marks a major breakthrough for Zhipu AI after launching the 100 billion base conversation model ChatGLM and ChatGLM2

ChatGLM3 is developed using an original multi-stage enhanced pre-training method. This method can make training more complete. According to the evaluation results, in 44 Chinese and English public data set tests, ChatGLM3 ranked first among domestic models of the same size. Zhang Peng, CEO of Zhipu AI, released new products at the press conference and demonstrated the latest product features in real time

ChatGLM3 new technology upgrade with higher performance and lower cost

ChatGLM3 launched by Zhipu AI has become more powerful with richer training data and better training solutions. Compared with ChatGLM2, MMLU increased by 36%, CEval increased by 33%, GSM8K increased by 179%, and BBH increased by 126%

At the same time, ChatGLM3 aims at GPT-4V and has implemented iterative upgrades of several new functions, including CogVLM with multi-modal understanding capabilities - image recognition semantics, which has achieved SOTA on more than 10 international standard image and text evaluation data sets. ; Code enhancement module Code Interpreter generates code and executes it according to user needs, automatically completing complex tasks such as data analysis and file processing; Web search enhancement WebGLM-access search enhancement can automatically find relevant information on the Internet based on questions and provide answers when answering Refer to relevant literature or article links. The semantic and logical capabilities of ChatGLM3 have been greatly enhanced.

ChatGLM3 also integrates the self-developed AgentTuning technology, which activates the model agent capabilities, especially in terms of intelligent planning and execution, which is 1000% improved compared to ChatGLM2; it also enables domestic large models to natively support tool calling, code execution, Complex scenarios such as games, database operations, knowledge graph search and reasoning, and operating systems.

In addition, ChatGLM3 this time launches end test models ChatGLM3-1.5B and ChatGLM3-3B that can be deployed on mobile phones, supporting a variety of mobile phones and vehicle platforms including vivo, Xiaomi, and Samsung, and even supporting CPU chips on mobile platforms. Inference speed can reach 20 tokens/s. In terms of accuracy, the performance of the 1.5B and 3B models is close to that of the ChatGLM2-6B model on public benchmarks.

Based on the latest efficient dynamic reasoning and memory optimization technology, the current reasoning framework of ChatGLM3 is better than the current best open source implementation under the same hardware and model conditions, including vLLM launched by the University of Berkeley and the latest version of Hugging Face TGI , the inference speed is increased by 2-3 times, and the inference cost is doubled, only 0.5 points per thousand tokens, the lowest cost.

This content is for reference only and does not constitute any investment advice. Readers should use their own judgment when using this information and assume responsibility for their own decisions. This website is not responsible for any losses caused by the use of this content

This account does not make any statement or guarantee as to the availability, accuracy, timeliness, validity or completeness of any information published, and hereby disclaims any liability or consequences that may arise from the information. After rewriting: This account makes no representation or warranty as to the availability, accuracy, timeliness, validity or completeness of any information posted, and disclaims any liability or consequences in this statement

2. This account is non-commercial and non-profit. The reproduced content does not mean that you agree with its views and are responsible for its authenticity, nor is it intended to constitute any other guidance. This website is not responsible for any direct or indirect responsibility for any inaccuracies or errors in any information reproduced or published.

3. The information, materials, text, pictures, etc. used in this article come from the Internet, and all reproduced content has been marked with the source. If you find any work that infringes your intellectual property rights or personal legal rights, please contact us and we will modify or delete it in a timely manner

The above is the detailed content of Zhipu AI launches the third-generation large base model ChatGLM3 to adapt to more domestic chips. For more information, please follow other related articles on the PHP Chinese website!