Hello, can someone say from experience how much faster custom LLM is than using 4o mini? How much was your latency improved and what are costs like? Also, what platform should I build on? I was looking into Phi 3...