I have some business automation process for some customers that I maintain, where Terra with low reasoning ended up with the best price/performance & latency. With an equivalent Luna with high reasoning, price was usually better, but quality had a higher variance (very much depends on where the thinking takes the model) and latency for high reasoning was worse.
I have some business automation process for some customers that I maintain, where Terra with low reasoning ended up with the best price/performance & latency. With an equivalent Luna with high reasoning, price was usually better, but quality had a higher variance (very much depends on where the thinking takes the model) and latency for high reasoning was worse.