Skip to content
The AI Agent Factory

Which model for which job

Approved by
Panaversity ·
11 min read

As verified on November 5, 2026.

Both apps let you pick a model, and both vendors give the same advice about it: the brief and the effort setting usually matter more than the model. Start with the model the vendor recommends for most work. Try a different model only when a job keeps failing its done-when at a higher effort setting, or when the job is routine, repeated often, or must be fast.1,2,3,4

The models are specialists, not rungs on one ladder. A faster model can beat a more capable one at its own kind of job, so compare models on your own tasks before settling. Three things are also easy to mix up:

  • The default: the model an app uses unless you choose another.
  • What you can select: the other models the app lets you pick.
  • What your plan includes: some models are available only on some plans, or use your limits faster.

The tables below say which is which.

ModelUse it forNotes
Opus 5.5Most delegated work: tasks that run for a while, reports and analysis, agentic coding, computer use, work with images and screensAnthropic's recommended starting point for most work. Default effort is medium
Sonnet 5.5Everyday work where speed matters: drafting, summaries, data analysis, routine coding, using toolsFast and capable, and lighter on your plan's limits
Haiku 5.5High-volume, simple, fast jobs: sorting and classifying, pulling fields out of documents, routing requests, helper sessions in a split jobThe fastest and cheapest. It also judges /goal conditions by default
Fable 5.1The hardest work: sessions that run for hours, deep research carried through to a finished document, spreadsheet or deck, and jobs where Opus 5.5 at higher effort still falls shortThe most capable model you can choose. It uses your limits much faster, and on some plans it runs only on paid credits5,4,6
ModelUse it forNotes
GPT-6 LunaQuick questions, short emails, summaries, triage, frequent automationsThe fastest and cheapest. Reported as the default in Chat on Free and Go
GPT-6.1 TerraScreen-based work and computer use, visual reasoning, concise writing, scoped code fixesA specialist: for its kind of job it can beat the larger models
GPT-6.1 SolComplex work where time and cost also matter: reports, analysis, comparing sources, debugging and code reviewReported as the default in Chat on paid plans
GPT-6 AstraAmbiguous problems, deep analysis and ambitious deliverablesOpenAI's most capable model. Reported as the model behind Work on Plus and higher, and selectable in Codex where your plan includes it. A pro mode does more work before answering, for when quality matters more than time7,2,8,9,10,11

By kind of job

The jobClaudeChatGPTWhat matters more than the model
A quick answer, a short email, a summarySonnet 5.5Luna or Sol in ChatA clear question
Writing you will send as yourselfOpus 5.5 or Sonnet 5.5Terra or SolExamples of your own writing
Research and reports from many sourcesOpus 5.5, or Fable 5.1 for the hardestWork, with Astra where availableNamed sources, and which one wins
Spreadsheets and numbersOpus 5.5Work, with Astra where availableCalculations done by tested code or formulas, never by the model's own arithmetic
A long task you leave runningOpus 5.5, or Fable 5.1 for hours-long sessionsWork, with Astra where availableA done-when the worker can check
A small, well-specified code fixSonnet 5.5Codex with Terra or SolA test that shows it works
A large change across many filesOpus 5.5, or Fable 5.1Codex, with Astra where availableTests, and review of the change
Operating a browser or your screenOpus 5.5Work, with Astra or Terra where availablePermission mode and approvals
Many small, repeated jobs, or helpers in a split jobHaiku 5.5LunaA checked sample of the results
A high-stakes decisionThe strongest model you haveThe strongest model you haveA person who reviews the evidence and signs off

Three cautions apply to any choice. A stronger model does not replace a human review: it makes better drafts, not accountable decisions. A model never does the final arithmetic in this book: calculations belong to tested code or formulas. And when you switch models, rerun your reusable briefs against their done-whens, because a brief tuned for one model may behave differently on another.

What matters more than any feature

Every feature on this page changes how work begins, where it runs and how long it can continue. None of them changes who is accountable for it. The book's first page, Claude and ChatGPT 2.0, sets out the four rules that matter more than any feature, and Part II teaches them: the Four-Part Brief, the Review Contract and the Authority Envelope.

Try it yourself. The practice lab on the first page has five short hands-on tasks, 10 to 20 minutes each, that you can do with a free account.

The Five Horizons comes next. It shows where this book is going, from one AI Worker to AI-native companies and countries.

Check yourself

Question 1 / 4

0 answered

Which Claude model does Anthropic recommend as the starting point for most work?

Sources

  1. Anthropic: "Tuning effort is often a better lever than switching models." OpenAI: compare settings "on representative tasks instead of assuming that the highest effort is always the best tradeoff." Using GPT-6, whose address still says gpt-5.6, OpenAI.↑
  2. Prompt guidance, OpenAI.↑↑2
  3. Models overview, Anthropic.↑
Owner
Panaversity

On this page