Robin Li: Baidu's ERNIE Model Inference Costs Reduced to 1%! Collaborations Established with Samsung, Honor, and Others
-
On February 29, during Baidu's Q4 and full-year 2023 earnings call, Robin Li, founder, chairman, and CEO of Baidu, revealed that the inference costs of Baidu's ERNIE large model have been reduced to 1%.
Li stated that since its release, Baidu has continuously reduced the inference costs of the ERNIE model, with the inference cost of the "ERNIE Bot" 3.5 version being 1% of that of the 3.0 version.
With the continuous reduction in inference costs, more and more companies are willing to test, develop, and iterate their applications on the "ERNIE Bot." Moreover, we believe there is still room to reduce costs in the future, making generative AI products more affordable for a broader user base, which will further enhance the adoption rate of 'Wenxin Yiyan'.
Robin Li also stated that the daily API calls for the Wenxin large model have exceeded 50 million, representing a 190% quarter-over-quarter growth. In December last year, approximately 26,000 enterprises utilized the Wenxin large model, marking a 150% quarter-over-quarter increase.
Currently, renowned companies such as Samsung, Honor, and Autohome have established partnerships with Baidu. We also hope to take this opportunity to build an ecosystem centered around 'Wenxin Yiyan' (ERNIE Bot), and we look forward to more and more users benefiting from it," said Robin Li.