Personally I think glm 5.3 flash (included in standard) and glm 5.3 (is the current quick assistant in standard), are better for most web search tasks, as they have a lower hallucination rate, which is more useful for most tasks than a slightly smarter model. Though I would encourage kagi to include more strong models in the standard plan. I'm currently strongly considering just paying for models and search (including kagi) over API, as at my current usage it would be a bit cheaper and a lot more flexible in model choice than what's currently available in the standard plan.
