The Verafy Swarm Explorer: One Question, Many Models
Compare answers across many language models and let models judge models
When you ask one chatbot a question, you get one polished paragraph and no way to tell what it left out. The Verafy Swarm Explorer asks many models at once and shows you the answers side by side.
Where it came from
Verafy began as a hackathon project. In December 2024 a demo called Truth Chain put model disagreement on screen, and it spread. About 4,000 people joined the community, and the project went on to a $17M capital launch and a Solana validator that peaked at 220,000 staked SOL. The Swarm Explorer is the product that came out of that demo.
How it works
- Many models answer. Type a statement or a question and choose how many models to ask. Each one answers on its own.
- Models judge models. Other models check the answers for factual accuracy and score them. You see each model's verdict and confidence, and an overall consensus.
- The split is the signal. Where models agree, you can be more confident. Where they split, that's usually where bias, missing context, or an outdated answer is hiding.
Each result card shows the question, the overall verdict, a confidence level, each model's score, and a short explanation with its source. Results can be shared, saved, and starred.
On your phone too
What's next
The Swarm Explorer now has two siblings. The Bias Detector runs a fixed set of 565 contentious questions against any model and gives each one a report card. The Fact-Checker for Chrome checks the claims on whatever page you're reading.
Try the Swarm Explorer at swarm.verafy.ai.