Yes. It does depend on the field though. It's tremendous with human languages in my experience. In mathematics and similar it often drifts into outright hallucination. I know the more recent models have been somewhat addressing this, but not sure of details. Fundamentally they don't "reason".