Infrastructure pricing.
Not model taxation.
Pay for the control plane, not your tokens. Bring your own API keys.
Calculate Your Potential Savings
See how cost-aware routing could reduce your LLM spend.
What this means: Many queries (summarization, classification, simple Q&A) produce equivalent results on cheaper models. Drag the slider to estimate what percentage of your traffic could route to lower-cost options.
- GPT-5 cost:$30/1M tokens
- Llama 3 (70B) cost:$0.5/1M tokens
- Your routable ratio:65% to cheaper models
Actual savings depend on your query mix, model choices, and provider pricing. This calculator provides estimates only.
Choose Your Deployment
Self-host for free, or let us manage the infrastructure in your region.
Self-Hosted
Deploy on your own infrastructure
Apache 2.0 License
- Full gateway functionality
- PII redaction engine
- Cost-aware routing
- Audit logging
- Deploy anywhere (Europe, on-prem)
- Community support via GitHub
- You manage infrastructure
Avarana Cloud
We run the infrastructure in your region
Pricing based on volume
- Managed European deployment
- Zero-touch infrastructure
- Smart routing included
- Enterprise SSO/SAML
- Dedicated support
- Custom SLAs available
- GDPR, UK DPA architecture
Sovereign
Private VPC with full control
Custom agreement
- Private VPC deployment
- Unlimited users & retention
- BYOK (Bring Your Own Keys)
- Custom routing policies
- Dedicated Support Engineer
- On-call incident response
- Architecture review included
Zero Token Markup. Bring your own OpenAI/Anthropic keys. We charge for infrastructure, not model access.
All options include European deployment, PII redaction, cost-aware routing, and audit logging.
Secure your infrastructure.
Tell us about your requirements. We'll scope the right solution.
Data Sovereignty First
Self-hostable in Europe, so your data never leaves your jurisdiction
Zero US subprocessors, unlike Portkey or hosted alternatives
Complete audit trail stored on infrastructure you control
Why Avarana? US-based AI gateways (including Portkey, now owned by Palo Alto Networks) cannot guarantee European data residency. Avarana is built from the ground up for jurisdictional compliance.
Common Questions
Do you charge per token?
No. You bring your own API keys (OpenAI, Anthropic, etc.). We charge only for the infrastructure. Zero markup on model usage.
Can I run this in my own VPC?
Yes. The Sovereign plan supports full Private VPC deployment. Your data never leaves your infrastructure. We provide Helm charts for Kubernetes.
How does smart routing save money?
Our routing engine analyzes prompt complexity and routes simple queries to cost-effective models (Llama 3, Claude Haiku) while reserving GPT-5 for complex tasks.