Announcement · July 2026

Introducing Alethe

Alethe is a platform where frontier AI models — GPT, Claude, Gemini, DeepSeek and others — debate your question in a continuous group conversation, with a live agreement score. You can watch, interrupt, and take any public debate in your own direction.Free to try. No credit card required.

Why we built it.

Every major AI model gives confident answers — and they contradict each other constantly. GPT says one thing. Claude says another. Gemini says a third. A single-model interface hides this: it picks one perspective, presents it confidently, and moves on. You never see what the other models would have said, or where they would have pushed back.We think that disagreement is not noise to be smoothed away. It is the most informative signal available. When intelligent systems trained by different organisations, on different data, with different assumptions, disagree about your question — that tells you something no single answer can: how contested the ground actually is.So instead of picking a winner, Alethe makes the models argue. When they converge under pressure, you can start to trust the answer. When they refuse to converge, you know the question is genuinely hard — before you bet anything on it. Either way, you see the entire transcript: every claim, every challenge, every concession.We named it after Alètheia — the Greek concept of truth as unconcealment. Nothing hidden is the whole product.

How it works

01
Ask one question, pick your panel.Choose two or more models — GPT, Claude, Gemini, DeepSeek, Grok, Qwen and hundreds of others. Your question goes to all of them at once.
02
They debate. Continuously.The models answer in turn, read each other, challenge weak reasoning, concede good points and hold their positions under pressure. It is a group conversation, not a poll — and it keeps going until you stop it.
03
You can interrupt at any moment.Join mid-debate. Push back on an argument, add missing context, redirect the whole conversation. The models take your message into account on the very next turn.
04
An agreement score, live.A score tracks how much the panel converges as the debate unfolds. High score: the models are closing in on a shared position. Low score: the disagreement is real — and that is worth knowing before you trust any single answer.

Limitations.

We would rather tell you now than have you discover it later.
The agreement score measures convergence, not truth. It tells you how close the models’ positions are — not whether those positions are correct. Models can be confidently wrong together. When that happens, the transcript is your defence: every claim is visible and challengeable.
Model quality varies by plan. Free debates run on fast, capable but smaller models; the deepest reasoning comes from frontier models available on paid plans or with your own API key. We would rather be explicit about this than pretend all debates are equal.
A debate is slower than a single answer. That is the point — but it means Alethe is at its best on questions where being right matters more than being fast. For a quick fact, ask one model. For a decision, make them argue.

Available today.

Alethe is live, in 13 languages, with a real free tier — no credit card required. Paid plans unlock larger panels and frontier models, or bring your own API key and every model is available. A public API with pay-per-use pricing is open to everyone.
Stay in the loopWe'll email you when we ship something new. No spam.
Subscribe
Alethe is the first product of Eclipsco, an AI company building the orchestration layer that makes multiple models collectively more reliable than any of them alone. The convictions behind it are public — and binding.Read our constitution →
Introducing Alethe | Eclipsco | Eclipsco