Pricing & Tiers
Transparent pricing, sovereign by default.
Pay for the models you run, in the infrastructure you control. No data egress fees, no per-seat lock-in, no hidden telemetry. Your models. Your cloud. Your control. Kervalt preserves that standard.
API Token Rates
Tafari Core
Per-token pricing on the Tafari Core family — billed per 1M tokens, input and output. Figures below are indicative.
Tafari Core S
Fast classification, extraction, routing.
- Input / 1M tokens
- €0.40
- Output / 1M tokens
- €1.20
Indicative pricing.
Tafari Core M
Balanced reasoning for production agents.
- Input / 1M tokens
- €1.50
- Output / 1M tokens
- €4.50
Indicative pricing.
Tafari Core L
Deep analysis and long-context synthesis.
- Input / 1M tokens
- €4.00
- Output / 1M tokens
- €12.00
Indicative pricing.
Zuberi Vector
Embeddings for sovereign retrieval (Qdrant).
€0.10 / 1M tokens
Sankofa Rank
Cross-encoder reranking for precision search.
€0.90 / 1M tokens
Deployment Tiers
Run where your compliance demands.
Same models, same control plane — deployed inside your VPC, in the Kervalt-managed Kanda Vault, or fully on-premises.
Your VPC
Custom
Token billing + your own infrastructure costs.
- Deployed inside your virtual private cloud
- Your keys, your network perimeter
- Zero data leaves your tenancy
Most Popular
Kanda Vault
from €3.500/month
Kervalt-managed, dedicated, EU-hosted.
- Instance-hour billing, dedicated capacity
- Min/max replica scaling under your policy
- Zero-data-retention guarantee
On-Premises
Custom
Licensed annually, air-gap capable.
- Your hardware, your data center
- Full model weights under your roof
- Offline operation supported
Enablement
Learning & Implementation Workshop
A 3-day interactive, hands-on program bridging DDD architecture with practical RAG deployment and secure agent orchestration in the Kazi Workplace. Sessions run 11:00–15:00 CET, limited to a maximum of 3 participants per cohort.
| Scope | Package A Kervalt Starter | Recommended Package B Enterprise Accelerator |
|---|---|---|
| Interactive Workshop | 3 days (11:00–15:00, max 3 participants) | 3 days (11:00–15:00, max 3 participants) |
| Sovereign EU-Hosting & Audit | Dedicated EU-server setup & zero-data-retention guarantee | Dedicated EU-server setup & zero-data-retention guarantee |
| Data Ingestion & Zuberi Vector (Qdrant) | Up to 2,500 documents (PDF, DOCX) | Up to 10,000 documents (incl. CAD/spec metadata) |
| Griot Agent Orchestration | 1 agent workflow (your choice) | 3 agent workflows (Checker, Matcher, Tender Assistant) |
| Quality Assurance & Evals | Basic ground-truth test set (5 questions) | Full ground-truth test set (15 questions) + parameter tuning |
| Methodology (DDD & Lean) | Basic RBAC & RAG configuration | Full DDD structure + 5S/Lean flow optimization |
| Post-Workshop Support | 14 days email | 30 days priority + 1-on-1 evaluation session |
| Investment (excl. VAT) | €4.500,- | €6.500,- |
Enterprise Billing
Procurement built for regulated enterprises.
Annual contracts with committed-use volume discounts, EU invoicing in EUR, and audit-ready documentation for your compliance and procurement teams.
Annual contracts
Predictable committed-use pricing with volume discounts at scale.
EU invoicing
EUR invoices from an EU entity, VAT-compliant, audit-ready.
Volume discounts
Tiered token and instance rates as your workloads grow.
Sovereignty, always
Every tier keeps data residency and control under your policy.
Ready to deploy sovereign AI?
Run Kervalt models in your own cloud, on your own terms.