GLM OCR is a focused specialist — it takes text as input and produces text as output, suggesting it's tuned for extracting and processing textual content from structured or semi-structured sources. As an open-weight model from zai-org, it trades broad generalist capability for targeted OCR-oriented behavior. Expect reliable text extraction within its narrow lane, but don't bring it problems outside that domain.