Some users like ChatGPT's new 30 configurations for task-specific control over speed, cost, and tokens, while many others call the options overwhelming and a needless complication that kills simplicity and auto-mode ease.
Based on 14 visible X reactions from 60 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@rasbt i actually like it. nice to have a few knobs to play with depending on the task and decide how much speed, cost, and tokens i want to throw at it
@rasbt That's analysis paralysis dressed as optionality. Auto mode died the moment every query needed a margin call.
@rasbt Look how much bundling and unbundling has accelerated! Clear sign of the singularity.
@rasbt agreed on this. Too many models that also have too many effort levels
Higher computational effort levels consume user limits faster.
@rasbt agreed on this. Too many models that also have too many effort levels
@rasbt 30 configs for one query is where i tap out
I like choices... but now I have: 2x modes (Codex vs. Work mode) 3x GPT-5.6 models (Sol, Terra, Luna) 5x effort levels (Light, Medium, High, Extra High, Ultra) That's 2 x 3 x 5 = 30 possible configurations for a query 🤯 What happened to "Auto" mode? https://x.com/rasbt/status/2075369179817902176/photo/1
I think it kind of makes most sense if we think of it as training vs inference-scaling (while keeping in mind that Sol, Terra, Luna are also different sizes) https://x.com/rasbt/status/2075407542042132825/photo/1
Also, is Ultra better than Pro or is it the same thing but different nomenclature? https://x.com/rasbt/status/2075385074757898490/photo/1
@rasbt There’s also fast mode, and number of subagents…. I’m sorry, we are working on consolidating this and make it easier to navigate
@rasbt Don’t forget. Normal. Fast. And cerebras.
Some users like ChatGPT's new 30 configurations for task-specific control over speed, cost, and tokens, while many others call the options overwhelming and a needless complication that kills simplicity and auto-mode ease.
Based on 14 visible X reactions from 60 accounts; directional sample.
Ask a question below.
Published answers will appear here.
I like choices... but now I have: 2x modes (Codex vs. Work mode) 3x GPT-5.6 models (Sol, Terra, Luna) 5x effort levels (Light, Medium, High, Extra High, Ultra) That's 2 x 3 x 5 = 30 possible configurations for a query 🤯 What happened to "Auto" mode? https://x.com/rasbt/status/2075369179817902176/photo/1