Baidu (NASDAQ:BIDU) said it released five new versions of AI model Ernie and an enhanced development toolkit for enterprises, as part of an upgrade to its large-scale model platform called Qianfan.
The new ERNIE models, which feature fewer parameters than large language models, or LLMs, are created to provide powerful performance across a range of applications with lower inference costs, the company said an e-mail to Seeking Alpha.
The models include: Ernie Speed, which specializes in understanding context, allowing it to generate more coherent and accurate content.
Ernie Lite, which is an upgraded version of Ernie Bot Turbo, offers 20% higher performance in tasks like sentiment analysis, multi-task learning and natural reasoning, and a 53% reduction in inference costs, according to Baidu.
Ernie Tiny is the smallest model among the three, and ideal for customers seeking affordability and low latency, mainly in applications such as retrieval, recommendation and intent recognition, the company noted.
Baidu AI Cloud also introduced two domain-specific models: Ernie Character — which is designed for scenarios based on roleplaying, such as allowing users to create non-playable characters, or NPCs, in gameplay or customer service.
Meanwhile, Ernie Functions is for customers needing integration of external tools or business functions. Enterprises can now skip additional fine-tuning and directly apply these proprietary models to develop their own smart assistants, the company added.
Baidu noted that Qianfan platform, with its ModelBuilder and AppBuilder, has served 80,000 enterprise users, helped customers fine-tuned 13,000 models, and developed 160,000 LLM-based applications.
The platform’s daily usage has increased by 97% quarter-over-quarter since December 2023, showing its role as a “super factory” for AI model development, as per the company.
