The wave of price cuts for large models sweeps the industry, Tongyi Qianwen combines Open Source and price reduction efforts.

Recently, Alibaba Cloud announced a significant reduction in the API call prices for its Tongyi Qianwen series of large models. Among them, the input price for the main model Qwen-Long dropped from 0.02 yuan per k tokens to 0.5 yuan per million tokens, a reduction of up to 97%. This move marks the beginning of a new round of price reductions in the large model industry.

In fact, Tongyi Qianwen is not the first large model provider to adopt a price reduction strategy. Since May, many well-known AI companies have launched low-priced products or reduced the prices of existing products:

• A new generation model API from an open source quantitative company is priced at about one percent of GPT-4 Turbo.
• An AI company has reduced the calling price of its personal version model by 80%.
• A well-known AI company has launched a new product with performance comparable to GPT-4 but at half the price.
• Two main large models under a search engine company announced full free access.
• A certain voice technology company announced that its API will be permanently open source.

Industry insiders analyze that this round of large-scale price cuts may stem from the advancements in large model inference technology and the decrease in costs. Price reductions help lower the barriers to entry for AI application developers, promoting the wider implementation of AI technology.

In addition to price reductions, Ali's Tongyi has also adopted an Open Source strategy. On May 9, Ali Tongyi released a series of large language models covering 500 million to 110 billion parameters to meet the needs of different scenarios:

• Small size model ( 0.5B-14B ) is suitable for deployment on mobile, PC, and other edge devices.
• The medium-sized model (32B) seeks a balance between performance and efficiency.
• Large-scale model (72B-110B) supports enterprise-level and research-level applications.

In addition, Tongyi has also open-sourced multimodal models such as vision, audio, and code. This series of initiatives aims to provide developers with more options and promote the diversified development of AI applications.

Industry experts say that "price reduction + Open Source" is becoming a consensus strategy among global large model manufacturers. This helps address two major pain points faced by AI application developers: high API prices and insufficient quality of open source models, thereby accelerating the implementation of AI applications. With the advancement of large model technology and the decline in costs, AI applications are expected to flourish in the future.