Guys I'm telling you the amount of utility you get from not having to count token or check quota is way larger than the marginal subjective capability difference between models. It's why I'm not too found of GLM so far even though it's very capable.Lmao, even Microsoft thinks closed-source is getting too expensive.