Alibaba has recently unveiled more than 100 open-source artificial intelligence models, significantly enhancing its proprietary technology to better compete with established rivals.
The newly released models, known as Qwen 2.5, are designed for various applications across sectors, including automotive, gaming, and scientific research. Alibaba claims these models exhibit advanced capabilities in mathematics and coding.
The company aims to rival domestic competitors like Baidu and Huawei, as well as American giants such as Microsoft and OpenAI.
Trained on vast amounts of data, Alibaba's models can understand prompts and generate text and images. The open-source nature allows anyone, including researchers, academics, and businesses worldwide, to utilize these models for creating their own generative AI applications, saving time and costs.
Alibaba hopes that this open-source approach will attract more users. The Chinese tech giant first introduced its Tongyi Qianwen or Qwen model last year and has since released improved versions, claiming its open-source models have been downloaded 40 million times to date.
Additionally, Alibaba has upgraded its proprietary flagship model, Qwen-Max, which is not open-source. Instead, Alibaba sells its capabilities to enterprises through its cloud computing products. The company asserts that Qwen Max 2.5-Max has outperformed competitors like Meta's Llama and OpenAI's GPT-4 in areas such as reasoning and language comprehension, though independent results are awaited.
Alibaba also introduced a new text-to-video conversion tool based on its AI models, allowing users to input prompts for the AI to generate videos, similar to OpenAI's Sora solution.