GLM-5.2 Caveman Ultra

GLM-5.2 Caveman Ultra is a text-to-text variant in the GLM family that pairs the upstream GLM-5.2 weights with an injected caveman output-compression ruleset. According to the EUrouter listing, the model shares the same underlying weights, context window, and per-token pricing as the base GLM-5.2, while the prompt steers it toward near answer-only replies. Caveman compression targets prose style exclusively, preserving code, commands, error strings, numbers, and API names verbatim, which makes the variant well suited to code-adjacent workflows where terse explanations matter more than fluent narrative.

Because the variant does not alter the core weights, its practical strengths mirror those of the broader GLM-5.2 lineage, including support for reasoning, tool calling, structured output, attachments, and temperature control, alongside an open-weights distribution. The aggregator listings describe it as a 753B-parameter model, positioning it as a large-scale general-purpose system within the family. Teams that want brief, command-style responses for scripting, debugging, or API orchestration tasks can lean on this variant, while those needing fuller prose can fall back to other GLM releases.