July 28, 2025:
Z.ai Unveils Cost-Efficient GLM-4.5 Model - Chinese startup Z.ai has launched the GLM-4.5 reasoning model, which is reportedly more cost-efficient than DeepSeeks R1. This model runs on affordable H20 graphics cards and offers lower token processing costs by leveraging its 355 billion parameters selectively. It surpasses several alternatives in AI benchmarks.
Additionally, a smaller version, GLM-4.5-Air, targets cost-sensitive applications. Despite U.S. export restrictions, Z.ai, backed by Alibaba and Tencent, aims to go public this year.
