In the 138-case comparison, GLM-5 solved 94 cases (64%) and MiniMax M2.7 solved 85 (57%); Claude Opus 4.6 solved 100 (68%), while GPT-5.4 solved 91 (61%). The suite spans file operations, tool use, retrieval, conversation, memory, summarization, and unit tests.
The result is specific to LangChain’s maintained evaluation suite and hosted inference configurations, not evidence that open models match frontier systems across all workloads. The post also presents illustrative price and latency comparisons and says the same evaluation suite can be run across model groups. Source: https://www.langchain.com/blog/open-models-have-crossed-a-threshold