GLM-4.7: Frontier intelligence at record speed – now available on Cerebras (www.cerebras.ai)

Z.ai has launched GLM-4.7, the latest model in the GLM family, now available on the Cerebras Inference Cloud. This upgraded model significantly enhances coding capabilities, tool-driven agents, and multi-turn reasoning, boasting performance advancements that make it the top open-weight model in real developer benchmarks. GLM-4.7 achieves high-quality code generation and editing faster than its predecessor, GLM-4.6, and rivals closed models like Claude Sonnet 4.5 in effectiveness—all while providing up to ten times better speed and price-performance. A key feature of GLM-4.7 is its ability to operate at real-time speeds on Cerebras' wafer-scale engine, generating code at around 1,000 tokens per second, which enables practical applications for live agents and coding assistants. Enhanced reasoning techniques, such as interleaved thinking and preserved thinking, allow for improved context retention and consistency during multi-step workflows. These changes not only reduce inference latency but also lower end-to-end costs and streamline user experiences, making GLM-4.7 a compelling choice for developers looking to deploy more intelligent AI solutions efficiently.