NativeAIHub

Effort Levels and Model Configuration

All plans (xhigh requires Opus 4.7 or 4.8)2 min read
LevelAvailable onBest forToken cost
LowLowAll modelsQuick lookups, simple edits, latency sensitive tasksLowest
MediumMediumAll modelsRoutine coding, cost sensitive workLow
High (default)HighAll models (default on Opus 4.8, 4.6, Sonnet 4.6)Most development work, balanced reasoningModerate
XhighXhighOpus 4.7 and 4.8 only (default on 4.7)Complex architecture, difficult debugging, autonomous workflowsHigh
MaxMaxAll models (session only)Extremely demanding tasks, may show diminishing returnsHighest

Effort levels persist across sessions (except max)

Low, medium, high, and xhigh persist across sessions once set. Max applies to the current session only. Ultracode also resets when you start a new session. If you set xhigh for a complex project, it stays at xhigh until you change it. Switching back to the default is as simple as running /effort auto.

Practical effort strategy

Start at the default (high) for most work. Switch to xhigh when you are tackling a difficult architectural problem, debugging a subtle bug, or running a long autonomous session that needs deep reasoning at every step. Drop to low or medium for batch operations where you are running many small, well-defined tasks and want to minimize cost. Use fast mode when you want speed regardless of effort level.