I don't want thinking models

actually from the OpenAI o3 publications, it seems the larger model has the function/tool capabilities. They likely are scared of it being used maliciously though. hence only the mini for now

likely it costs money and thinking models are new harder way of doing things. if anything, it is probably more cost effective to have a non-thinking model handle tool functions and have a tool for it to query a thinking model. it is possible adding too much tool calling into the thinking model can degrade its capabilities for thinking. its a mixed bag of unknowns. surprisingly a lot of these LLM companies are running on bare minimum margins. GPT pro's 200/m doesnt even make money for openai most of the time. Claude 3.7 does work, so that is something you can use if you wanted to. probably one of the better LLMs out there for tool calling, coding or other non-creative stuff. don't know what else to say, its big enough mystery to warrant curiosity but not enough to hate on thinking models.