GLM-5.1 Tops Open-Source Models in Coding Agent Benchmark

GLM-5.1 has emerged as the leading open-source model in the Artificial Analysis Coding Agent Benchmark, according to a report by Artificial Analysis. The benchmark evaluates model performance on three key tests: SWE-Bench-Pro-Hard-AA, Terminal-Bench v2, and SWE-Atlas-QnA, which simulate real-world programming and technical tasks. While the proprietary Opus 4.7 model secured the top global position, GLM-5.1, operating on Claude Code, led among open-source models, showcasing its advanced capabilities in programming agent scenarios.