Question 1

What is Autopilot?

Accepted Answer

Autopilot is the default way to ask on AI Crucible. It reads your question - what kind it is, how hard, what is at stake, and whether it needs the web or files - and plans the run: one model for questions it reads as simple and low-stakes, or a panel of specialists on different models with a lead reviewer for harder or high-stakes ones. You see the plan and its estimated cost before anything runs, and you can open it in Advanced mode to change any part.

Question 2

What do the speed and cost settings change?

Accepted Answer

Speed sets how many rounds a panel may take: one at Fast, up to two at Balanced and up to three at Thorough. A panel stops early once it agrees, except at Thorough with Max quality, which runs every round. Cost sets the panel size and the models behind it: the smallest panels on fast, low-cost models at Lean, frontier models at Standard, and the strongest models, with a larger panel on hard questions, at Max quality. Autopilot chooses everything else. It may use stronger models when it is unsure how to read a question, and scales a run down if your balance cannot cover it; the plan shows both before you run.

Question 3

What is a team on AI Crucible?

Accepted Answer

A team is a pre-configured panel of AI specialists for one field. AI Crucible ships five teams: Medical, Legal, Financial, Construction, and a General team for everything else. Each has its own roles - an analyst, a researcher, a reviewer, a writer and an adversarial challenger - and a lead reviewer who weighs the panel and signs off on the answer. Every specialist runs on a different AI model, so the panel is genuinely several models arguing rather than one model playing parts.

Question 4

What does the effort level change?

Accepted Answer

Effort is one dial that sets three things: how many specialists sit on the panel, how many rounds they discuss for, and which quality band of models they are drawn from. Economy uses three specialists for one round on fast models. Standard uses five for two rounds on mid-tier models and turns on web lookups. Premium uses six, including an adversarial challenger, for three rounds on the strongest models available, with reasoning effort turned up.

Question 5

How many questions do I get on each plan?

Accepted Answer

Plans are measured in tokens rather than questions, because a question at Economy costs a fraction of one at Premium. The effort selector shows how many questions your plan buys at each level, computed from the models a team actually resolves to and their current prices, so the figure moves when models or prices do. A Starter plan buys roughly a thousand Economy questions a month, or about a dozen Standard ones.

Question 6

Do I have to use Autopilot or a team?

Accepted Answer

No. Autopilot calls a field team on its own for high-stakes questions in that field, and you can pin a team yourself in Teams mode. Advanced mode is the full workbench: pick the models yourself, choose any of the ensemble strategies, write your own role briefs, set the number of rounds and the arbiter. Opening an Autopilot plan or a team in Advanced pre-fills it, so you can start from either and then change anything.

Question 7

Which AI models do the teams use?

Accepted Answer

Models are chosen at run time from whichever are currently available, matched to each role by capability and to the effort level by quality band. Nothing is pinned to a specific model, so when a provider ships a new model or retires an old one the teams pick it up without any change to the team itself.

Question 8

Is a team answer professional advice?

Accepted Answer

No. The Medical, Legal and Financial teams each carry a disclaimer for exactly this reason. They are research aids that show their reasoning and where the specialists disagreed, so you can check the work - not a substitute for a clinician, a lawyer or a licensed advisor.

Question 9

What is convergence detection?

Accepted Answer

Convergence detection automatically stops the panel early when the specialists have effectively agreed, rather than paying for another round that changes nothing. It saves 10-30% on the cost of a run and is on by default at the Economy and Standard effort levels.

Ask once. Autopilot builds the right AI team.

Two dials: speed and cost

Speed

Cost

Teams Autopilot calls on

Pin a team and choose how hard it works

How it works

Watch a real panel at work

Why a panel beats one model

Case studies

Simple, transparent pricing

Build your own

Explore