After price reductions in May and September of this year, Alibaba Cloud has once again announced a price cut for its large models, marking the third round of price adjustments this year. The price reduction is significant, with the Tongyi Qianwen series visual understanding models seeing an overall decrease of over 80%. The price of the Qwen-VL-Plus model has dropped by 81%, with an input cost of only 0.0015 yuan per thousand tokens, setting a record low price online; the higher-performance Qwen-VL-Max has been reduced to 0.003 yuan per thousand tokens, a decrease of 85%. According to the new pricing, 1 yuan can process approximately 600 images at 720P or 1700 images at 480P.
The Qwen-VL series large models are multi-modal models launched by Alibaba Cloud, which have become one of the most popular models in the open-source community, boasting powerful visual reasoning capabilities. This model can not only recognize images of various resolutions and aspect ratios but also understand long videos exceeding 20 minutes and possesses the ability to visually comprehend tasks performed by intelligent agents such as mobile phones and robots. Qwen-VL is widely used in various visual recognition scenarios across devices, including smartphones and automobiles.
The Alibaba Cloud Bailian team stated that this price reduction is primarily due to the continuous optimization of Alibaba Cloud's infrastructure and model architecture, as well as the economies of scale brought about by the exponential growth in model usage. With ongoing technological advancements and optimizations, Alibaba Cloud's inference efficiency has significantly improved. The elastic AI computing power scheduling system built by Alibaba Cloud, combined with the Bailian distributed inference acceleration engine, has not only greatly reduced model inference costs but also sped up inference times. Alibaba Cloud also mentioned that as the visual understanding capabilities of Qwen-VL continue to improve, this model has become one of the fastest-growing models on the Bailian platform.
To further reduce the costs for users utilizing the large model API, Alibaba Cloud Bailian has introduced a new KV Cache billing model. This model significantly lowers model invocation costs by automatically caching context to avoid redundant computations, making it particularly suitable for scenarios such as long texts, code completion, multi-turn dialogues, and specific text summarization.
As Alibaba Cloud continues to optimize its infrastructure and models, the price reduction of the Qwen-VL series visual understanding models not only makes AI technology more accessible but also brings more application opportunities for developers and enterprises. By continually enhancing performance and lowering usage costs, Alibaba Cloud is promoting the widespread adoption and application of AI technology, providing stronger technical support for the digital transformation of various industries.
