Z.ai Launches Cost-Effective GLM-5.3-Flash AI Model Using Chinese Chips

Z.ai has introduced GLM-5.3-Flash, an open-source artificial intelligence model featuring 320 billion parameters. According to The Decoder, this new model performs nearly as well as the larger GLM-5.3, trailing by just three points on Artificial Analysis's Intelligence Index while costing only a seventh as much.

In a notable shift from typical AI deployments, The Decoder reported that all inference operations for GLM-5.3-Flash run exclusively on Chinese AI chips rather than Nvidia hardware. This approach underscores efforts to reduce dependency on foreign semiconductor technology and lower operational expenses.

For Japanese investors and tech companies, this development highlights growing competition in the AI hardware space and potential cost efficiencies, which could influence local strategies in AI adoption and chip sourcing amid evolving global supply chains.