GLM-5.2 Caveman

GLM-5.2 Caveman is a compression-oriented variant of GLM-5.2, sharing the same upstream model while applying built-in instructions that favor shorter answers. It is intended for workflows where output length matters but explanations still need to retain technical substance, especially when preserving code, commands, error strings, numbers, API names, and important warnings.

The model’s practical distinction is stylistic compression rather than a separate output filter: rules direct it to retain validation, error handling, authentication, and content explicitly requested by the user. This makes it a fit for concise explanations, technical assistance, and applications that need lower verbosity without intentionally sacrificing critical implementation details. The compression is instruction-based, so shorter output is a strong tendency rather than a strict guarantee.