ewpireUse cases · Consensus
Compare model answers
Ask one question to several models, read every model's answer, and get a synthesis of where they differ and where they agree.
The problem
Models differ on the same question, and one chat hides it. Before you rely on an answer – a technical choice, a figure, a claim you plan to publish – you want to see the spread.
How it works
- 01
Open Consensus, write the question and choose a format.
- 02
Pick the models to compare. They can come from different labs.
- 03
Each model answers independently, then critiques the others' answers without knowing whose they are.
- 04
Read the synthesis, then open each model's answer to see who said what.
What you give it · What you get
What you give it
- One question, as specific as you can make it
- Optional files for the models to read
- The models to compare
What you get
- The synthesized answer
- Model agreement and areas of disagreement
- Unique insights: what only one model raised
- Each model's own answer and critique
What to know
- Agreement is not proof: models can be wrong even when they agree.
- A confidence figure is shown for the General, Coding and Decision formats. The other formats are not meant to agree, so none is shown.
The prompt the button opens
Which database should a two-person team pick for a new SaaS product that expects 10,000 users in its first year: PostgreSQL, MySQL or MongoDB? Compare them on operations effort, hiring, migration risk and cost, and say where you are least sure.
Try it with your own words
The button opens Consensus with this prompt in the box, for you to change. Signed out, you sign in first and land there with the prompt kept.