GLM 5.1

GLM-5.1 is presented as a next-generation flagship large language model (LLM) specifically designed for agentic engineering. It boasts significantly enhanced coding capabilities compared to its predecessor, GLM-5. The model demonstrates state-of-the-art performance on benchmarks like SWE-Bench Pro and shows considerable improvements on NL2Repo (repository generation) and Terminal-Bench 2.0 (real-world terminal tasks). A key differentiator of GLM-5.1 is its ability to sustain optimization over much longer horizons in agentic tasks. Unlike previous models that tend to plateau after initial gains, GLM-5.1 is built to remain effective over extended sessions. It approaches complex problems by breaking them down, conducting experiments, interpreting results, and precisely identifying blockers. Through repeated iteration, GLM-5.1 revisits its reasoning and refines its strategy, allowing it to maintain optimization over hundreds of rounds and thousands of tool calls. This iterative and adaptive approach means that the model's performance improves the longer it operates on a task, handling ambiguous problems with superior judgment.