How it worksFrom a question to collective truth — the complete journey of an Alethe AI debate.

You ask a questionType your question, problem, or topic. No specific format required — one sentence is enough. Alethe AI adapts to all types: simple questions, complex topics, philosophical debates, technical problems.E.g. "Should AI be regulated?" or simply "What is the capital of Australia?"

Models debate in real timeThe models you selected take turns speaking. Each one reads the full conversation transcript before responding — it sees everything the others said. You watch each AI write live.Each model argues, can contradict the previous ones, or converge when agreement is clear.

Models keep going until your limitThe conversation runs in round-robin — each model responds in turn, indefinitely. It stops when your plan's token limit is hit (Free: 30K, Pro: 80K, Ultra: 200K, Apex: 500K) or when you stop it manually.All messages are saved — you can resume the conversation later.

Stop on your termsYou can stop the conversation at any time. Your plan's token limit is the only automatic stop — it ensures the conversation stays within your allocated resources.No score-based stop, no arbitrator. You stay in control.

You get the truthThe result is a conversation built collectively by multiple intelligences — more reliable than a single model's response. Exportable as PDF (Ultra+).No single-model bias. Disagreements are visible, convergences are real.

Why is it more reliable?

A single model can hallucinate — multiple models reading and contradicting each other, much less.

Each model reads what the others wrote — no isolated parallel responses.

You see the disagreement in real time — you know exactly where the AIs diverge.

More features

📎Attach a file or image

You can attach a document (PDF, DOCX, XLSX, TXT…) or an image (JPG, PNG, WEBP) to your message. Content is extracted client-side and passed to the models. Images are analyzed by Claude Vision before the debate.

🎤Dictate with your voice

Press the mic button to dictate your question. Real-time transcription is powered by Deepgram (nova-2). Language is auto-detected based on your language preference.

🎭Choose a specialized mode

Before launching a debate, pick from 7 modes: Alethe (default), Friendly, Business, Wellness, Socratic, Debate, or Custom (your own prompt). Each mode changes how the AIs behave — not the models themselves.

Agreement Score

Conversation modes