The standard free ChatGPT LLM you get after a few messages HAS to be some sub-20b model with online search enabled, no other way to explain how awful it is

Recently bought into the local LLM hype by buying a 32gb vram gpu and holy shit, gemma 4 31b at 5bits blows the standard ChatGPT model out of the fucking water. I just can't unsee the quality difference now that I've experienced it. Does Openai just cut costs by serving their customers this slop sin

中文全文翻译正在补齐，可先查看下方英文原文。

Recently bought into the local LLM hype by buying a 32gb vram gpu and holy shit, gemma 4 31b at 5bits blows the standard ChatGPT model out of the fucking water. I just can't unsee the quality difference now that I've experienced it. Does Openai just cut costs by serving their customers this slop sin