Alibaba has released more than 100 open Qwen 2.5 AI models, as well as an AI text-to-video converter

Chinese technology giant Alibaba announced the release of a total of more than a hundred new open-source artificial intelligence (AI) models, and also introduced AI technology for converting text to video. Reuters writes about this.

Image source: starline/freepik.com

The new open-source models are part of the Qwen 2.5 family, Alibaba’s large core language model released in May of this year. The presented models have different parameters and are tailored for different areas of application. The company said the new models have from 0.5 to 72 billion parameters, offering knowledge of mathematics, programming and support for more than 29 languages. The models are intended for use in various sectors of the economy, including automotive, gaming and scientific research.

Unlike competitors such as Baidu and OpenAI, which stick to closed-source AI application development, Alibaba has adopted a hybrid model, investing in both proprietary and open-source development to expand its AI product portfolio.

The company also said it has updated its own flagship model called Qwen-Max, which is not open source. Instead, Alibaba sells its capabilities to businesses through its cloud computing products. Alibaba said Qwen Max 2.5-Max outperformed competitors such as Meta✴ Llama and OpenAI’s GPT4 in several areas, including reasoning and language understanding.

Alibaba also introduced a new AI model for converting text to video, belonging to the Tongyi Wanxiang family of AI tools for generating images from text prompts. This puts Alibaba in direct competition with global players such as OpenAI, which is also working in this direction, as well as local companies such as ByteDance (owns TikTok), which launched Jimeng AI, a text-to-video AI application in August, on Apple App Store for Chinese users.