Skip to content
Multilo Docs

Choosing a model for your task

How to pick the right model by capability tier and task type — with sample prompts — without memorizing model names that change over time.

You don’t need to memorize model names to choose well. Multilo groups every model into a small set of capability tiers; pick the tier that fits the task, and the model picker shows which current models sit in each tier. Because the tiers stay constant even as the underlying models change, this guidance never goes out of date.

How to think about it

Two questions decide the right tier for any task:

  1. How hard is the reasoning? Reformatting a paragraph is easy; critiquing a methodology or synthesizing ten papers is hard.
  2. How much does speed and cost matter? A quick pass you run twenty times a day should be cheap and fast; a final-draft pass can afford the most capable model.

Easy + high-volume → a faster, cheaper tier. Hard + high-stakes → the most capable tier. Most everyday work sits comfortably in the middle.

The three tiers

Multilo sorts models into three tiers by capability. The picker labels which models are in each.

TierBest forTrade-off
FastHigh-volume, low-complexity work: formatting, summaries, quick edits, simple explanations.Cheapest and quickest; less depth on hard reasoning.
BalancedThe everyday workhorse: drafting, rewriting, citation work, and most coding and analysis.Strong reasoning at a moderate cost per run.
FlagshipThe hardest tasks: methodology critique, multi-source synthesis, long autonomous drafts, tricky proofs and code.Most capable; costs the most credits and runs slower.

Match the model to the task

TaskStart with
Formatting, cleanup, short summariesFast
Everyday drafting & explaining a conceptFast → Balanced
Citation formatting & switching stylesBalanced
Inline suggestions while you writeFast
Claim Check citation verificationBalanced
Methodology critique, argument, synthesisFlagship
Full Draft Writer (whole document)Flagship for quality · Balanced to save credits
Math, proofs, or complex codeFlagship

When in doubt, start lower and step up

Run a Balanced model first; if the result isn’t deep enough, re-run the same step on Flagship. You’ll spend fewer credits and only reach for the most capable model when the task actually needs it.

Sample prompts by task

Each prompt below is tagged with a good starting tier. The point isn’t the exact wording — it’s matching the difficulty of the ask to the tier.

  • Fast — tidy-ups and summaries:
    “Tighten this paragraph without changing its meaning.”
    “Summarize this section in three bullet points.”
  • Balanced — everyday drafting and citations:
    “Draft a 200-word introduction for a section on survey-design bias, citing my library.”
    “Reformat every citation in this document to APA 7.”
  • Flagship — deep reasoning and synthesis:
    “Critique the methodology in this section: identify threats to validity and suggest concrete fixes.”
    “Synthesize where these five sources agree and disagree on remote-work productivity, and note the strongest counter-argument.”

MODES: a second lever

Model tier is one dial; an agent’s MODES are another. MODES change how thorough an agent is — a light pass versus a deep one — independently of which model it uses. For a demanding task you can pair a Flagship model with a deeper MODE; for a quick pass, a Fast model with a light MODE. They multiply, so tune both to the stakes.

Cost & plan access

More capable tiers cost more credits per run, because they use more expensive models. Fast models are available on the free plan; the most capable tiers are typically premium and need a paid plan. Matching the tier to the task is also how you keep credit use efficient — don’t spend a Flagship run on a task a Fast model handles well.

Where the model names live

You’ll notice this guide never names a specific model. That’s deliberate: the models on offer change as providers release new versions, but the tiers don’t. The model picker in the app is the single source of truth — it shows the current models in each tier and flags which are premium. Pick the tier this guide recommends, and choose whichever model the picker offers there.

Per agent and per run

Each agent has a sensible default model, and you can override the model for a single run from the picker — so you can keep a fast default and reach for Flagship only on the runs that need it. See Models.