Tencent Cloud Liu Yuhong talks about the low penetration rate of large models: still in the development stage, AI product solutions are immature

According to news from this website on May 30, Liu Weihong, vice president of Tencent Cloud and head of Tencent Hunyuan Large Model, said that the current low penetration rate of large model storage is due to two main reasons: on the one hand, large models are still in the development stage; AI product solutions are immature; on the other hand, user awareness is still being established, and the demand for AI is focused on search/efficiency.

Liu Yuhong said that the original search engine may have done a good job, and the current large model can make the search better. Since Tencent released its large model product in September last year, model efficiency has increased by 50%, from 100 billion parameters to 1 trillion parameters, and supports 256k native window context and 16-second video generation. At present, more than 600 businesses within Tencent have accessed the Hunyuan large model.

On May 22, 2021, Tencent Cloud released a new large model price plan. One of the main models is the Hunyuan lite model. The total API input and output length is planned to be upgraded from the current 4k to 256k, and the price is from 0.008 yuan. /Thousand tokens are adjusted to be fully free.

Hunyuan-standard API input price dropped from 0.01 yuan/thousand tokens to 0.0045 yuan/thousand tokens, a decrease of 55%, and API output price dropped from 0.01 yuan/thousand tokens to 0.0045 yuan/thousand tokens. 0.01 yuan/thousand tokens dropped to 0.005 yuan/thousand tokens, a decrease of 50%. The newly launched Hunyuan-standard-256k has the ability to process ultra-long texts of more than 380,000 characters. The API input price is reduced to 0.015 yuan/thousand tokens, a decrease of 87.5%, and the API output price is reduced to 0.06 yuan/thousand tokens, a decrease of 50%.

The highest configured trillion parameter model Hunyuan-pro, the API input price dropped from 0.1 yuan/thousand tokens to 0.03 yuan/thousand tokens, a drop of 70%.

The above price adjustments take effect immediately. At the same time, in order to better serve developers, Tencent Yuanqi, a one-stop intelligent agent creation and distribution platform, has also comprehensively upgraded its model resource support plan: users can not only distribute exclusive AI agents to Tencent’s global channels with one click, but also receive free Hunyuan large model resources and traffic support can also be used to distribute and deploy agents in the form of APIs to various scenarios. The free token limit has been increased from the previous 1 million to 100 million.

The above is the detailed content of Tencent Cloud Liu Yuhong talks about the low penetration rate of large models: still in the development stage, AI product solutions are immature. For more information, please follow other related articles on the PHP Chinese website!