Simple pricing for every use case
Local inference is always free. Cloud Pro adds 130+ premium models with one credit system — chat, OCR, vision, audio, and image all share the same pool.
1 chat message · 1 OCR page · 1 vision query · 1 minute of audio · 1 image- Unlimited local models
- 10 cloud chats / day
- 5 OCR pages / day
- 3 vision queries / day
- No audio, no image gen
- Community support
- Everything in Free
- 100 chats / day (~3000/mo)
- 30 OCR pages / day
- 50 vision queries / mo
- 20 min audio (STT/TTS)
- 10 image generations / mo
- Everything in Starter
- 500 chats / day (~15k/mo)
- 100 OCR pages / day
- 500 vision queries / mo
- 200 min audio / mo
- 200 images / mo
- 30 min Realtime Voice / mo
- Priority cloud routing
- Everything in Pro
- Shared usage pool
- 50k chats / month / seat
- 10k OCR pages / month / seat
- SSO & SAML
- Audit log
- Centralized billing
- Everything in Team
- SLA 99.9%
- Self-hosted registry
- Dedicated GPU quota
- Custom contracts
- BAA, ISO 27001, GDPR
What you get in each tier
| Feature | Free | Starter | Pro | Team |
|---|---|---|---|---|
| Local LLM inference (any size) | ✓ unlimited | ✓ unlimited | ✓ unlimited | ✓ unlimited |
| Local vision, STT, TTS, image gen | ✓ unlimited | ✓ unlimited | ✓ unlimited | ✓ unlimited |
| Chats per day (cloud) | 10 | 100 | 500 | shared pool |
| OCR pages per day (powered by german-ocr.de) | 5 | 30 | 100 | shared pool |
| Vision queries / month | ~90 | 50 | 500 | 5000 |
| Audio (STT + TTS) minutes / month | — | 20 | 200 | 1000 |
| Realtime voice mode | — | — | 30 min | 300 min |
| Image generation | — | 10/mo | 200/mo | unlimited |
| Priority cloud routing | — | — | ✓ | ✓ |
| Private model library | — | ✓ | ✓ | ✓ |
| Team admin / SSO / audit | — | — | — | ✓ |
Frequently asked
Is local inference really unlimited?
Yes. Pull any model from the Hub — Llama, Qwen, DeepSeek, Mixtral, Whisper, Piper, FLUX, anything — and run it on your own machine forever, without an account, without telemetry on your prompts. The paid tiers only matter if you want to use the Cloud features for models that are too big for your hardware.
What is a "credit"?
One credit equals one cloud action: a chat message, an OCR page, a vision query, a minute of audio, or one generated image. We don't charge per-token because nobody knows how many tokens are in a token. Credits are the same currency for every model type so you can mix and match.
How do you keep cloud cheap?
We route to Google Vertex (Gemini Flash) for general cloud chat/vision/audio, to our own german-ocr.de service for OCR, and to Together / OpenRouter / our own GPU pool for open-weight inference. We pay wholesale, you pay retail, the difference covers infrastructure plus a thin margin.
Can I evaluate Pro commercially before buying?
Yes. You may run DeepNetz commercially for evaluation purposes for up to 30 days without a Pro subscription, even at a company. After 30 days you need a license to keep using it commercially. See the EULA.
What about academic and non-profit use?
Academic research, classroom teaching and non-profit use are all covered by the Free tier license — no purchase required.
Refunds?
14 days, no questions asked.