well I'm using mini models and GPT-5.4-mini is better than Gemini 3.1 Pro (considerably faster in exchange for less broad world knowledge) and 3.5 Flash (that one's trash by itself)
You are, sure, but I wonder how many weren't due to the name. In many people's minds, it's like, why even bother with a non frontier model? Especially if they're not paying for it, via their employers' subscriptions. This new OpenAI naming changes it though, they equate Sol to Fable so people will use Terra instead which has roughly 5.5 performance, thereby saving inference for OpenAI.