The training data likely references Claude significantly more often than Kimi, given the popularity of the models. There will simply be more examples of “Claude” being the response to that question.