Claude Opus 4.8 has one setting most people will ignore.

Effort level.

That small dropdown decides how much Claude should think, how fast it should answer, and how much of your usage it burns.

The simple version:

Low → quick answers

Medium → normal work

High → safe default

xHigh → serious coding and agentic work

Max → deep, expensive thinking

Ultra Code → xHigh + Dynamic Workflows inside Claude Code

_

This is how I would use it:

Low

For quick questions, summaries, formatting, small rewrites. Cheap. Fast. Around 10% of High on simple tasks.

Medium

For emails, outlines, basic plans, content ideas. Good when the answer needs shape, but not deep research. Around 40% of High.

High

This is the Opus 4.8 default. Use it for serious writing, customer-facing work, document analysis, code review, policy thinking, and anything you do not want to redo. This is your 100% baseline.

Max

Only use it when mistakes are expensive. Architecture. Legal review. Risk analysis. Hard debugging. Deep research. It can cost 4x to 8x more than High depending on the task.

Ultra Code

This is not normal chat mode. It is Claude Code running bigger workflows with subagents. Use it for codebase audits, multi-file refactors, security checks, test generation, migrations.

Not for small fixes.

_

Claude Code commands:

/effort low

/effort medium

/effort high

/effort xhigh

/effort max

/effort ultracode

The mistake is using Max for everything.

That feels powerful, but it is usually just waste.

The better move is routing.

Small task → Low

Normal task → Medium

Important task → High

Hard coding task → xHigh

High-risk thinking → Max

Big engineering workflow → Ultra Code

That is the real Opus 4.8 upgrade.

Not just a smarter model.

A better way to spend thinking.