Products

One platform. Every intelligence.

Alethe AI is the only platform where GPT, Claude, Gemini and 20+ other models debate each other in real time — converging toward a single, more reliable answer than any model could produce alone.
35AI models
7Debate modes
Up to 500KTokens / debate

35 models, one conversation

Every major provider. Swap models in one click. Each model reads what the others wrote — this is a real debate, not a side-by-side comparison.
GPT-5 NanoOpenAI
Gemini 2.5 FlashGoogle
DeepSeek V3.2DeepSeek
GPT-4.1 MiniOpenAI
Claude Haiku 4.5Anthropic
Llama 4 MaverickMeta
Qwen 3.6 PlusAlibaba
Gemini 3 FlashGoogle
GPT-5OpenAI
GPT-4.1OpenAI
Grok 3 MinixAI
Mistral MediumMistral
Claude Sonnet 4.6Anthropic
Claude Sonnet 4.5Anthropic
Gemini 3.1 ProGoogle
GPT-5.4OpenAI
Grok 3xAI
DeepSeek R1DeepSeek
Claude Opus 4.6Anthropic
GPT-5.4 ProOpenAI
o3 ProOpenAI
Grok 4xAI

Built for reliability, not engagement

Every feature exists to help you get a better answer. No infinite scroll. No gamification. The product's job is to be right.
Real-time streamingWatch models write their responses character by character, simultaneously. No waiting for the full answer before reading.
Semantic agreement scoringA live score (0–100%) updates after every message — measuring how much models actually agree using vector embeddings, not keyword overlap.
Automatic stop conditionsThe debate stops on token cap or when you choose to stop. The current model always finishes its sentence before halting — no abrupt cutoffs.
File & image uploadAttach PDFs, DOCX, Excel, images. Content is extracted and shared with all models. Images are analyzed by Claude Vision.
Voice inputDictate your question via microphone. Transcription via Deepgram nova-2 — low latency, high accuracy, 13 languages.
Message injectionSend a follow-up while models are still debating. The debate adapts mid-stream — no need to restart from scratch.
Prompt cachingKV caching reduces token costs by 60–70% on long debates. The longer the conversation, the cheaper each subsequent message.
Stop at any timeA single click gracefully stops the current model mid-debate without cutting it off mid-sentence. Clean, predictable exit.

7 conversation modes

One platform, seven different ways to use it. Switch modes between debates — the underlying engine is the same, the behavior adapts completely.
Alethe
Analytical multi-AI discussion. Debates when the question calls for it, answers naturally otherwise. The default mode.
Models read each other's responses and challenge positions in real time.
Friendly
Warm, pedagogical, accessible. Perfect for learning or exploring a topic casually without overwhelming technical depth.
Uses simpler language, more examples, patient pacing.
Business
Strategic advice, ROI focus, numbered decisions. Built for the boardroom — structured outputs, action items, trade-offs.
Models argue the business case, play devil's advocate, then converge.
Wellness
Practical guidance on nutrition, sleep, stress and healthy habits — always recommending a professional for medical concerns.
Focused on lifestyle and prevention, not medical diagnosis.
Socratic
Questions only. The AI guides your reasoning without giving you the answer — forces you to think, not just receive.
Powerful for learning, interview prep, and philosophical exploration.
Debate
Models take opposing positions and argue them hard. Designed for intellectual sparring, not polite convergence.
Exposes the strongest arguments on all sides of a contested question.
Custom
Your system prompt, your rules. Define exactly how every model should behave, what to focus on, and what to avoid.
Evaluated and scored by our AI before the debate starts.

Simple, transparent pricing

Start free. Upgrade when you need more power. Save 20% on annual plans.
Free
$0forever
Explore multi-model debates at no cost.
8 debates / month
2 models max
30K tokens / debate
Alethe & Friendly modes
Tier 1 models
Get started
Pro
$9/mo
More power for curious minds.
20 debates / month
3 models max
80K tokens / debate
PDF export
Business & Wellness modes
Tier 1–2 models
Start Pro
Most popular
Ultra
$34.99/mo
For researchers and power users.
20 debates / month
5 models max
200K tokens / debate
PDF export
All 7 modes
Tier 1–3 models
File & image upload
Start Ultra
Apex
$99/mo
Full platform access for teams and builders.
40 debates / month
Unlimited models
500K tokens / debate
Public API access
24h priority support
All 35 models
Custom mode
Start Apex
20% discount on all annual plans. No contracts. Cancel any time.

Full plan comparison

Feature
Free
Pro
Ultra
Apex
Debates / month
8
20
20
40
Max models / debate
2
3
5
Unlimited
Max tokens / debate
30K
80K
200K
500K
PDF export
All 7 modes
2 modes
4 modes
6 modes
Tier 3 models
Tier 4 models (Frontier)
Custom mode
Public API access
Priority support (24h)

Public API

Apex plan
Integrate multi-model debates directly into your product. Submit a question, poll for the result, and surface a consensus answer inside your own UI — without building the orchestration layer yourself.

Authentication

Generate an API key from your account settings. Pass it on every request as a header. Keys are hashed and never stored in plain text.
X-API-Key: apex_xxxxxxxxxxxxxxxxxxxxxxxx

Endpoints

GET
/api/v1/models
Returns the list of available models and which tier each belongs to.
curl https://fast-api.eclipxco.com/api/v1/models \
  -H "X-API-Key: apex_xxxxxxxxxxxxxxxxxxxxxxxx"

// Response 200
{
  "models": [
    { "slug": "openai/gpt-4.1-mini", "name": "GPT-4.1 Mini", "tier": 1 },
    { "slug": "anthropic/claude-sonnet-4-6", "name": "Claude Sonnet 4.6", "tier": 3 },
    { "slug": "google/gemini-2.5-flash", "name": "Gemini 2.5 Flash", "tier": 1 }
  ]
}
POST
/api/v1/debates
Start a new debate. Returns immediately with status 202 and a polling URL — the debate runs asynchronously.
curl -X POST https://fast-api.eclipxco.com/api/v1/debates \
  -H "X-API-Key: apex_xxxxxxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "What is the most reliable approach to LLM evaluation?",
    "models": ["openai/gpt-4.1-mini", "anthropic/claude-sonnet-4-6"],
    "mode": "super_intelligence"
  }'

// Response 202
{
  "debate_id": "dbt_8fKpXm2wQn",
  "status": "running",
  "poll_url": "/api/v1/debates/dbt_8fKpXm2wQn"
}
GET
/api/v1/debates/{id}
Poll for results. When status is done, the full transcript and agreement score are available.
curl https://fast-api.eclipxco.com/api/v1/debates/dbt_8fKpXm2wQn \
  -H "X-API-Key: apex_xxxxxxxxxxxxxxxxxxxxxxxx"

// Response 200 — while running
{ "status": "running", "messages_count": 4 }

// Response 200 — when complete
{
  "status": "done",
  "agreement_score": 0.83,
  "messages": [
    { "model": "GPT-4.1 Mini", "text": "..." },
    { "model": "Claude Sonnet 4.6", "text": "..." }
  ],
  "stop_reason": "token_cap"
}

Async flow

POST /debatesSubmit question + models
202 + poll_urlRuns in background
GET /debates/{id}Poll every 2–3s
status: doneRead transcript + score
Rate limit: 60 requests per minute. API keys are managed in Account settings — available to Apex subscribers only.

Frequently asked questions

How is this different from just asking ChatGPT?
When you ask ChatGPT a question, you get one model's answer — shaped by its training, its biases, its blind spots. Alethe AI puts multiple models in the same conversation. They read each other's responses, challenge positions, correct errors, and build on each other's reasoning. The result is a more complete, more stress-tested answer than any single model could produce alone.
What does "debate" actually mean here?
It's not a side-by-side comparison. Each model reads what the others wrote before composing its response — then agrees, disagrees, or refines. It's a real conversation, with positions that evolve across messages. Some questions produce quick consensus. Others surface genuine disagreements between models that are worth knowing about.
How long does a debate take?
Most debates complete in 30 to 90 seconds depending on the number of models, the complexity of the question, and your plan's token limit. Responses stream in real time — you start reading as the first model writes, not after everything is done.
Is my data private? Are my conversations stored?
Yes, conversations are stored to power your history and the Explorer feature. They are private by default — only you can see them. You can publish a debate to make it publicly visible, but that's always an explicit action. We do not use your conversations to train any model.
Can I share or export my debates?
Yes — two ways. You can publish any completed debate to a public link and share it with anyone (no account required to read). You can also export the full transcript as a PDF — available on Ultra and Apex plans.
What happens when I hit my token limit?
The current model finishes its sentence, then the debate ends cleanly. You get a notification with your usage. Unused tokens do not carry over to the next month — limits reset on your billing date.
Is there a free plan? Do I need a credit card?
The Free plan is free forever — not a trial. No credit card required. You get 8 debates per month, 2 models per debate, and access to the Alethe and Friendly modes. Upgrade only when you need more.

Start debating today.

8 free debates a month, no credit card required. The free plan is free forever — not a trial.