Agent Atlas

Atlas · building

Model and effort

Choosing a model and how much effort it spends. Two Claude Code builders publish when they go low, medium or high.

Emerging practiceDepth 2 of 4Reviewed 2026-11-12

In plain words

Claude Code lets you choose a model and adjust its effort level. Higher effort asks the model to spend more compute. The skill is spending it where it counts.

How it works

Anthropic’s overview says to start with Opus 5.5 for most work. Boris Cherny says Opus 5.5 works at any effort level, that low trades intelligence for speed, and that he uses medium about 95 percent of the time. Thariq Shihipar implements on low and verifies on high. Paweł Huryn hands a second look to a different model than the one that wrote the code.

Going deeper

These are a few builders’ published habits from May to October 2026, not rules. The Builders page has them with sources, and a budget picker for one $20 plan.

Sources

  1. Models overview Anthropic
    • If you’re unsure which model to use, start with Claude Opus 5.5 for most workloads.
    • Use Claude Fable 5.2 for demanding reasoning and long-horizon agentic work, or when your evals on Claude Opus 5.5 at higher effort still fall short.
    • Reliable knowledge cutoff
    • If you’re unsure which model to use, start with Claude Opus 5.5 for most workloads. Use Claude Fable 5.2 for demanding reasoning and long-horizon agentic work
    • Claude API ID claude-fable-5-2 claude-opus-5-5 claude-sonnet-5-5 claude-haiku-5-5
    • $8 / input MTok
    • $40 / output MTok
    • $4 / input MTok
    • $20 / output MTok
    • From $0.10 / input MTok
    • From $0.50 / output MTok
    • Context window 1M tokens 1M tokens 1M tokens 1M tokens
    • For high-volume, latency-sensitive tasks such as classification, extraction, and routing
  2. Opus 5.5 at low, medium, and xhigh Boris Cherny on X
    • A cool thing about Opus 5.5 is it is nice to use at *any* effort level.
    • If you want faster speed and lower intelligence use low;
    • if you want more thinking and intelligence use xhigh.
    • I use medium for most work.
  3. Medium most of the time, higher for complex visual work Boris Cherny on X
    • For complex visual stuff with Opus xhigh/max, but 95% of the time I stick to medium. The best default is different for each model
  4. Using Claude Code: Spending your effort Thariq Shihipar
    • implementing on low effort, reviewing what it built, and then running verification on high effort.
    • Low: for when I want quick responses that are in the loop, e.g. brainstorming, sketching, easy changes
    • Medium: for most of my regular software engineering work, e.g. new feature implementation.
    • High: for work where verification is important or there are edge cases, e.g. fixing a bug in a brownfield codebase.
    • Max: When I want Claude to operate fully autonomously to solve difficult problems
    • Try varying effort for Opus 5.5 and Fable 5.1 based on your task or even mid-conversation by using /effort in Claude Code
  5. How to Set Up Codex as a PM and Run It Next to Claude Paweł Huryn
    • I hand it to the runtime that didn’t write it. Different model, different misses.

Quotes were checked against the live pages on 2026-11-14.