53 Why I Don't Recommend Using Inferior Models? Here's a real case I encountered. Yesterday, I ran Gemini all day. After Pro ran out, I switched to Flash. Emm, after 200 million tokens, there was no progress, and it even broke the previously working pipeline. Finally, with GPT 5.4, it took 2 hours to quickly fix the issues and complete the task in one go. Inferior models try very hard. How hard? You'll see. I thought after saying this, even the dumbest AI would make some progress, right? However, the follow-up to this was that Gemini detected a dead loop, and then waited for
Purchase required to continue
This is a paid article. After signing in, your purchase will be unlocked automatically.
No comments yet. Be the first to share your thoughts.