Liangjiang enterprise unveils advanced AI computing platform

A pioneering company in Liangjiang New Area has launched a state-of-the-art AI public computing power sharing platform, now open for commercial use. This innovative platform provides computing power at a competitive rate, priced at just 90 percent of the market standard.

The platform, with an impressive capacity of 224 PFlops, is designed to meet the diverse AI needs of small and medium-sized enterprises and research institutions. It supports a wide range of applications, including AI model training, AI application development, and foundational scientific research.

The platform's developers and operators have introduced a Model as a Service (MaaS) model, known as TokenPony. This service aims to deliver high-performance, cost-effective large model solutions to local businesses.

Enterprises can access hundreds of mainstream model services via API, including full-featured versions of DeepSeek-R1 & V3, Qwen, and Kimi. This eliminates the complexities associated with GPU selection, model downloading, environment configuration, code debugging, and deployment.

The platform is equipped with an ultra-large context window of up to 164K, resolving issues of context fragmentation in long document processing, complex multi-turn dialogues, and intricate agent tasks. It also supports automatic failover and load balancing, ensuring uninterrupted business operations.