Also on the team was the Chinese-American “AI Godmother” Li Feifei.
The Exploration of the “cheapest” AI Model
This comes after China’s Deepseek surprised the world by releasing a high-performance, cost-effective open-source AI model.
The Chinese AI startup left many surprised by the claim that its chatbot was built at some cost of what was developed by the American tech giant.
This has raised questions across the billions of dollars spent by US-based tech companies expanding data centers that are thought to be necessary to unlock the next wave of AI.
What Qwen Research says
According to the research paper, the results were obtained after the inference model was trained with solutions to 1,000 curated questions. The S1 model was also created to undergo a “thinking process” distilled from the Gemini Thinking Experimental Model developed by Google.s. On January 29th this year, Alibaba announced a new version of QWEN 2.5. The AI model claimed to be superior to DeepSeek-V3 and Meta’s Llama-3.1-405b.
The QWEN 2.5 series was first announced in September. According to the South China Morning Post, the series ranged in sizes from 500 million to 72 billion parameters.
Low cost
To develop an inference model for the S1, we spent just $14 for the cost of running a graphics processing unit (GPU). This is based on the computer described during the study and states that it was trained for 26 minutes on 16 Nvidia H100s.
Tips can be rented for $2 per hour.
Pan Jiayi, a computer scientist at the University of California, Berkeley, said the basic model is the key to training such a low-cost inference model. “The quality of the basic model is important,” he pointed out.
The QWEN2.5 series was announced last September by Alibaba’s cloud computing unit.
FAQ
1. Who is the competitor for the S1 inference model?
The major competitor of the S1 AI model is the Openai O1 rival.
2. Is the newer version of Qwen 2.5 better than its rivals?
At launch, Alibaba claimed that the latest Qwen 2.5 would outperform the DeepSeek-V3 and Llama-3.1-405b.
Disclaimer Statement: This content is written by a third party. The views expressed here are those of the respective authors/organizations and do not represent views from the economic era (ET). ET is not responsible for, warrants, approves, or in any way. Take all the steps required to ensure that the information and content provided is correct, updated and verified. ET disclaims all warranties, express or implied, relating to the Report and the content therein.
