Cloud features and model usage are separate.
A subscription unlocks Sofia Cloud features. Inference is metered separately at the model’s published rate — bring your own key and Sofia’s charge is zero.
Free
For individual developers who want the harness without a cloud dependency.
- Sofia CLI
- Sofia Desktop
- Local models
- BYOK providers
- Local tools + MCP
Pro
Cloud conveniences for an individual without locking model choice behind the plan.
- Everything in Free
- Sofia PAYG models
- Cloud API keys
- Hosted capabilities
- Included monthly model credit
Team
Shared models, policy, capabilities and billing for software teams.
- Everything in Pro
- Teams + model grants
- Budgets + concurrency
- Shared provider keys
- Central billing + audit
Enterprise
Identity, provisioning, policy, support, and deployment requirements for larger organizations.
- OIDC + SAML SSO
- SCIM users + groups
- Desktop policy
- Postpaid billing
- Private model endpoints
Hosted inference, server tools, and remote execution are usage-based. Model rates vary by cache state, modality, batch mode, and provider time window.
Exact meters, visible rates.
Sofia records the price-book version used for every request, then stores immutable line items for the actual meters the provider reported.
| Usage path | Sofia subscription | Sofia inference charge | Who pays provider |
|---|---|---|---|
| Local model | Not required | $0 | You / local infrastructure |
| BYOK | Not required | $0 | Your provider account |
| Sofia PAYG | Not required | Published model rate | Sofia |
| Team managed key | Team | $0 | Organization provider account |
| Sofia-hosted model | Optional | Published Sofia rate | Sofia |