When to Use Each aiDex Mode: Solo, Compare, Judge, Pipeline, Team

Every mode in aiDex exists because one question shape rewards a different way of talking to models. The pillar piece on multi-model AI workflows covers the why. This piece covers the when.

What does each mode do, in one line?

In aiDex, the mode you pick decides how many models speak, in what order, and whether anyone gets the final say.

• Solo sends your prompt to one model and renders the answer.
• Compare sends the same prompt to 2 to 4 models in parallel and lays the answers side by side.
• Judge runs your prompt through several models and asks a lightweight moderator to pick (or merge) the best answer.
• Pipeline passes a task through fixed stages (Draft, Critique, Revision, Polishing), each handled by the model you assign.
• Team is an open conversation: up to five models, a moderator deciding who speaks next, and you steering the room.

Use your own provider keys or the ones we manage, and pick the models you want. Every mode lives behind the same chat panel; switching is one click.

When should I use Solo?

Use Solo when you want a fast, single answer and you already trust the model for the task.

Good Solo work: drafting a short message, asking for a quick definition, getting a one-shot code snippet, running a known prompt against your preferred model. Solo costs the least per question because only one model fires.

Skip Solo when the answer carries real stakes (publishing, decisions, contracts). One model is one opinion. The other modes exist to keep that opinion honest.

When should I use Compare?

Use Compare when you want to see how 2 to 4 models answer the same prompt and pick the answer that fits.

Compare shines for prompts where style matters as much as substance: copy variants, code that has to read a certain way, explanations targeted at a specific audience. Run the prompt once, read the columns, copy the winner. Our guide on how to compare AI models walks through the flow.

Compare is also the right call when you do not yet know which model is best for a new task. Two or three side-by-side runs answer that question faster than reading benchmarks.

When should I use Judge?

Use Judge when you want the models to vote and a moderator to land on one final answer.

Judge is Compare plus a verdict. Three models answer, a lightweight moderator reads the three answers, points out where they agree and disagree, and surfaces the shared answer as simple accept or reject cards. Useful for fact-style questions, plan reviews, and anything where consensus matters more than personality. The deeper how-to lives in how to get AI consensus.

Skip Judge when you specifically want the models to argue (use Team instead) or when one model is clearly better for the job (use Solo).

When should I use Pipeline?

Use Pipeline when the work has stages and each stage rewards a different model.

A Pipeline runs through four roles: Draft, Critique, Revision, Polishing. Assign a model to each stage. The draft model produces V1. The critique model attacks it. The revision model rewrites in response. The polishing model edits for cadence. You get a finished artefact instead of a chat log. The pattern is laid out step by step in how to build an AI pipeline.

Pipeline is the best mode for: long-form writing, code that needs both ambition and discipline, important emails, contract clauses, anything that benefits from "draft and then push back on the draft."

When should I use Team?

Use Team when the question is open and you want to watch three or four models talk it over.

Team is the panel format. Up to five models around the table, the moderator deciding who speaks next, you in the middle steering. Use it for strategy questions ("should we raise our prices next quarter?"), planning sessions, design reviews, anything where you want disagreement on the record. The full walkthrough sits in how to create an AI team.

Team costs more per question than the other modes because every model is in the room for every turn. Use it where the conversation itself is the artefact.

Which mode for which kind of work?

A short decision table covering most real questions:

Almost any task can run in more than one mode. Pick the smallest mode that matches the stakes: cheaper if the task is small, fuller if it is not. When in doubt, start in Compare; switching modes is one click. For where each mode fits into the bigger picture, the pillar piece multi-model AI workflows ties them together.