Alibaba Intelligent Information Business Group today officially released the full-stack self-developed, 100-billion-level parameter Quark model.According to reports, in the CMMLU authoritative large model performance evaluation, the Quark large model ranked first. According to the evaluation,The overall capability of the Quark large model has exceeded GPT-3.5, and is better than GPT-4 in some scenarios such as writing and exams.
The Quark large model is a self-developed multi-modal large model based on the Transformer architecture. It trains and fine-tunes hundreds of millions of graphic and text data every day. It has the characteristics of low cost, high response, and strong comprehensive capabilities.
At the same time, the Quark large model will also derive vertical models such as general education, medical care, and education, and can provide professional services such as AIGC and intelligent retrieval.
In the domestic professional examinations, Quark's large model scored close to full marks in the college entrance examination, and passed the clinical practitioner qualification examination with a high score of 486. This is a veritable "student master".
In terms of application,Quark's large model will be used in general search, medical and health, education and learning, workplace and other scenarios. Quark App will also be fully upgraded with the help of this large model.