Yu Panβs review basically says the quiet part out loud: OpenAI is the thing you ship with right now. He calls out a real jump in quality, faster responses, and an ecosystem that actually lets devs and teams build production workflows instead of cute demos. Compared to Gemini 2.5 Pro, he still picks OpenAI because the APIs, docs, and community feel like a long-term platform, not just a clever model drop.
But itβs not a fanboy post. He wants more transparency around model behavior, clearer best practices, better long-term project memory, saner pricing as usage grows, and real answers on stuff like Realtime Voice reliability and image OCR for hard cases.
If youβre using OpenAI, add your own review: where is it rock solid, where does it break, and what would make you actually trust it for the next few years?