[TechWeb] On May 21, after Alibaba Cloud announced a 97% price reduction for its Tongyi Qianwen GPT-4 main model this morning, Baidu just announced that its two major models, ERNIE Speed and ERNIE Lite, are free.
ERNIE Speed and ERNIE Lite are the "lightweight" models officially launched by Baidu in March this year. The "lightweight" large model can also achieve the effect of the general large model even using the small-scale expert model, while the reasoning cost of the small-scale expert model will be lower and cheaper.
The ERNIE Speed reasoning scenario has a maximum context length of 128k. When processing tasks such as knowledge questions and answers, it can better handle context dependencies and generate more coherent and accurate predictions or answers.
Compared with ERNIE Speed, ERNIE Lite has smaller parameters, which is more suitable for processing reasoning tasks on the AI accelerator card with low computing power. While taking into account the model effect and reasoning performance, it significantly reduces the cost of customer landing applications.
Since this year, the price competition in the domestic AI large model industry has become increasingly fierce.
Last week, the main model of bean curd under Byte Beat was priced as low as 0.0008 yuan/thousand tokens in the enterprise market, and the price of DeepSeek-V2 reached 1 yuan per million tokens. Previously, Zhipu AI also adjusted the price of glm3 turbo to 1 yuan per million tokens.
This morning, Alibaba Cloud announced that nine major AI models, including Qwen Turbo, Qwen Plus, and Qwen Long, the main GPT-4 models of Tongyi Qianwen, announced a comprehensive price reduction. Among them, the input price of Qwen Long API dropped from 0.02 yuan/thousand tokens to 0.0005 yuan/thousand tokens, with a maximum drop of 97%. (appropriate month)
