What changed
Amp’s smart, deep, rush and large modes were replaced by a Dial with four levels. At launch, ultra ran Claude Fable 5, high and medium ran GPT-5.6 Sol at different reasoning efforts, and low ran Z.ai’s open-weight GLM-5.2, each with a second oracle model for reviews. Reasoning effort became part of each level, removing the separate effort toggle. Amp later removed the Dial again on November 12.
Who it affects
Amp users whose prompts or habits were built around the old mode names.
What to do
Move workflows that named the old modes to a Dial level, keeping in mind Amp has since switched to a per-thread model selector.