Two-Path Pricing for Developers
Understand when queries are free and when they cost money.
Key takeaway: The two-path model means most users pay only the flat license. External providers are opt-in, not default.
Path 1: Browser-Native Generation (Free)
- The grounded generation engine runs entirely in the user's browser via the browser runtime and GPU acceleration.
- No API call is made to any server.
- No Stripe meter event is generated.
- Cost: covered by the flat Pauhu® license fee (or free tier: 3 queries/day).
- The user's query never leaves their device.
- Works offline after first model download.
How it works
- User types a query in pauhu.ai.
- Browser loads optimized model from OPFS cache (or downloads once from CDN).
- Multilingual embeddings generate query embedding locally.
- Embedding matches against local semantic index snapshot.
- The grounded generation engine produces a grounded answer.
- Result displayed - zero network calls.
Path 2: External Provider (Metered)
- Activated when the user explicitly requests an external AI provider (OpenAI, Anthropic, etc.).
- Requires explicit consent for third-country data transfer (GDPR Art. 49(1)(a)).
- Each request generates a Stripe meter event.
- Cost: transparent per-request pricing.
- The API gateway routes the query to the selected provider's API.
How it works
- User selects an external provider in settings.
- Consent dialog appears (third-country data transfer disclosure).
- User confirms consent.
- Query is sent via the API gateway to the external provider.
- Stripe meter event is recorded.
- Response is returned to user.
- Monthly invoice includes all metered external calls.
Pricing Comparison
| Path | Cost | Meter Event | Data Location | Offline |
|---|---|---|---|---|
| Browser-Native Generation | Flat license / Free tier | No | User's device only | Yes |
| External Provider | Per-request pricing | Yes (Stripe) | Provider's servers (US/EU) | No |
For API Consumers
- API calls to
/v1/searchand/iate/*endpoints are metered per your API key tier. - API calls are always server-side (Path 2 pricing model applies).
- Free tier: 3 requests/day.
- Paid tiers: see pricing page.
Note: The two-path model means most users pay only the flat license. External providers are opt-in, not default.
Why This Model Exists
At GTC 2025, Jensen Huang observed that inference costs are falling rapidly. Token-based pricing - pay per query, per token, per minute - is built on a cost structure that is compressing. Pauhu® is built for what comes after.
Instead of metering compute, Pauhu sells access to pre-scored, annotated data. Three flat tiers - Live Feeds, Annotated, and AI Training Data Access - give you 4.8 million documents across 20 EU institutional sources with legal modality classifications, topic codes, and cross-references. You pay for the depth of annotation, not the volume of queries.
The AI that reads this data runs in your browser. Optimized models are cached locally. Queries stay on your device. There is no token meter and no usage bill at the end of the month.
For tasks that exceed local capability - long-form generation, vision, code execution - Pauhu passes through to external providers at their published rates, with no markup. This is opt-in and clearly separated from the data license.
The data is the product.
Questions? Contact us at support@pauhu.eu
Full documentation: Documentation index