KERVALT / MODELS
One place to find the right model for the job.
Run advanced, open, and specialized models on European infrastructure. Compare speed, capacity, and response targets side by side.
01 / MODEL CATALOG
Model catalog
All models run behind your data perimeter. Your inputs don't leak to outside services, and they aren't used to train third-party models.
| Model type | Use case | Capacity | Speed target |
|---|---|---|---|
| Advanced reasoning | Hard analysis, coding, and step-by-step thinking | 128K–1M tokens | First answer in under 500 ms |
| Open general | Document Q&A, summarization, and sorting | 32K–128K tokens | First answer in under 200 ms |
| Specialized | Legal, financial, clinical, and engineering tasks | 32K–256K tokens | First answer in under 150 ms |
| Lookup and retrieval | Tafari search and matching | Variable | Answers in under 50 ms |
02 / WHY CHOOSE KERVALT
Why teams choose Kervalt for models
Deploy, benchmark, and fine-tune models without sending your data to outside services.
No lock-in
Run models on your own bare-metal nodes. Switch model types without rebuilding your app.
Built-in benchmarking
Compare accuracy, speed, and cost across models with reproducible evaluation suites.
Secure fine-tuning
Train adapters on your own data without sending inputs to outside services or public model providers.
NEXT STEP
Find the right model for your work
Talk to our solutions team about benchmarking, fine-tuning, and deployment options.