Key information
- Alibaba Cloud has reduced the price of its advanced Large Language Models (LLM) to attract Chinese businesses.
- The Qwen-VL-Max model now costs 0.003 yuan (0.00039 euros) per thousand tokens, which is lower than the prices charged by competitors such as ByteDance Ltd.
- More than 90,000 Chinese companies have downloaded Alibaba Cloud’s Qwen models, amid fierce competition from other Chinese companies.
Price reduction
Alibaba Cloud has significantly reduced the price of access to its most advanced Large Language Models (LLM), in a bid to attract Chinese businesses. The move was announced via WeChat and first reported by the South China Morning Post.
The company’s Qwen-VL-Max model, a visual reasoning model capable of understanding both text and images, now costs just 0.003 yuan (0.00039 euros) per thousand tokens. This price is significantly lower than competitors such as ByteDance Ltd, the parent company of TikTok.
Competition in the AI market
This aggressive pricing strategy reflects the intense competition between China’s leading technology companies to secure a dominant position in the burgeoning AI market. Alibaba Cloud faces rivals such as Tencent Holdings Ltd, Baidu Inc, JD.Com Inc, Huawei Technologies Co. Ltd. and ByteDance, all of which have launched competing LLMs in the past 18 months.
LLMs are AI models trained on massive data sets to generate human-like responses to user questions and prompts. They form the basis of various generative AI applications, including chatbots such as Google’s Gemini and OpenAI’s ChatGPT, next-generation search engines such as Perplexity AI, and image generators such as DALL-E.
Market development
Alibaba Cloud is targeting Chinese companies that see the potential of generative AI to improve productivity, like their counterparts in the United States. The company said more than 90,000 businesses in China have downloaded its Qwen templates.
However, Alibaba faces fierce competition. Over the past year, Chinese generative AI developers have released more than 250 new LLMs for public use. The country is home not only to major tech companies, but also to an impressive number of promising start-ups, including DeepSeek. Recently, DeepSeek introduced the DeepSeek-V3 LLM, an open-source model with 671 billion parameters, making it one of the most powerful models in its class.
The Qwen model range
Alibaba Cloud’s Qwen model family includes Qwen-VL, Qwen-VL-Chat, Qwen-VL-Max, Qwen2-VL, and the experimental QVQ-72B-Preview model. Qwen2-VL-Max excels in benchmarks such as DocVQA and MathVista, outperforming models such as OpenAI’s GPT-4V and Google’s Gemini Ultra.
Price reductions
This aggressive pricing strategy is not new for Alibaba Cloud. In February, it announced price reductions of up to 55 percent on several core cloud computing services, followed by a reduction of up to 97 percent on its original Qwen-VL model in May.
If you want access to all articles, subscribe here!
