At a Glance · AI-generated

Z.ai released GLM-5.3-Flash, an open-source multimodal mixture-of-experts model capable of processing 1 million tokens in context. The 320 billion active parameter model, also known as Ox Alpha, delivers performance comparable to leading competitors while significantly reducing computational costs and operating without Nvidia hardware. The release prompted immediate comparisons with Z.ai's standard GLM-5.3 variant, with benchmarks showing competitive performance across coding and routing tasks.