Arpia Credits

At ARPIA, we believe in adding measurable value to your organization. To achieve this, we have designed a credit consumption system that is transparent and focused on pricing based on value generation.

Credits are consumed based on items and value units. The following sections explain these items and their credit costs.

---

Credit Groups

---

Arpia Value Units (VUs)

If your account is billed using Value Units (VUs), this section applies.

Value Units are the default billing method. They ensure you are charged only when measurable value is generated by the platform.

You can analyze usage details in the Workarea Usage Page, including where value is generated and its associated costs.

VUs are calculated based on the resources and features used in workareas.

Every 1 VU = 1 USD (American Dollars).

---

Oria Commercial Packaging & Pricing

Three levels, billed per seat per day, so a seat added mid-month costs only what it used. Billing is prorated daily and invoiced monthly, consistent with the Workshop & Platform VU billing method above. A user added mid-month and removed before month-end is billed only for the days they were actually present on the account; there is no full-month charge for partial-month usage. The AI itself is paid separately through the AI Wallet, and you set the ceiling.

The line between the three plans is how far what you build can travel: Basic builds applications that live in the user's own Oria, Pro builds Team Business Apps that can be shared with a team, and Builder changes the platform everyone else runs on.

All three plans are prorated daily: users are charged only for the days they were active in the account during the billing month.

Application access is governed, not universal. Basic and Pro users see and run the applications a Builder has granted them by profile. Access is an administrative decision, not a plan entitlement — upgrading a seat does not by itself widen which applications that seat can see.

Oria Machine

Oria Machine is a governed coding environment. It lets Oria delegate multi-step coding tasks — reconciliations, bulk validations, scripts, data transformations — to a dedicated execution environment instead of attempting them inline in a session.

• Background execution. Long-running work continues without holding the session open.
• Human-in-the-loop approvals. Steps that change state are surfaced for approval before they run.
• Scoped access. A Machine reaches only the nodes, MCP endpoints and tools available to the session that launched it — nothing wider.
• Never touches platform source. Oria Machine executes work; it does not modify the ARPIA platform itself.

The more capable Reasoner (ORA) runs on Oria Machine, which is why it begins at Pro rather than Basic. Pro seats may create up to 3 Oria Machines. Users needing platform-scope build capacity beyond that belong on Arpia Builder.

Experts (coming soon)

Pro users will be able to share Experts with their team, alongside Team Business Apps. Sharing scope and governance follow the same profile-based access model described above. Availability and any effect on pricing will be published when the capability ships.

AI Wallet (usage-based, billed separately)

All AI usage on the platform — on any plan, whether generated by a Basic session, a Pro Reasoner run, or an application a Builder deployed — draws from the AI Wallet. The AI Wallet is a shared, company-wide prepaid consumption pool, sized and billed according to actual usage, and is not subject to daily proration since it is metered by consumption, not seat-days. It is not included in the fixed plan price.

Because the Wallet is company-wide, consumption generated by an application follows the organization, not the individual seat that happened to run it. Use hard limits and budget controls to set the ceiling.

BYOM / BYOK (Bring Your Own Model / Bring Your Own Keys)

Clients may connect their own model provider accounts or API keys instead of using ARPIA-provisioned models. In this case, ARPIA does not charge for the underlying token cost (billed directly by the provider to the client), but a flat $1.00 per 1 Million tokens, input and output combined, governance fee applies. This fee is a flat pass-through charge for AI governance, budgeting, model assignment, and audit, independent of provider and independent of model capability or intelligence grade.

Example: a client using their own OpenAI or Anthropic key processes 50M tokens (input + output combined) in a month. ARPIA charges $50 in governance fees for that usage, regardless of provider or model tier.

Positioning: Basic drives enterprise AI adoption (governed consumption and in-session building); Pro serves individual power users whose tools other people end up needing; Arpia Builder serves platform builders and governance owners; AI Wallet governs AI consumption cost.

Workshop & Platform VUs

---

DataApps VUs

If a feature is not listed above, its cost is 0 VUs.

---

⚠️ Notice — AI Solutions & Token Consumption

Value Units cover platform orchestration only for AI-driven features (AI Worker / AI Operator, AI Assistant). Token Consumption generated by these solutions when using a Shared LLM Model is billed separately under LLMs Core Models @ Arpia Cloud, at the selected model's published input and output rates per 1 Million Tokens.

• Shared LLM Models (no BYM): billed per the Core / Frontier model rates (Credits per 1 Million Tokens, input and output metered separately).
• BYM (Bring Your Model): no ARPIA token charge — only VUs (or Computing Hours) apply.

Token Consumption scales with the size and complexity of each interaction and the model selected. Monitor consumption in the usage dashboard and configure alerts for both Core and Frontier models.

🔒 Data retention: Prompts and completions sent to shared LLM providers are handled per our Privacy Policy — LLMs & AI Tools, including provider zero-/short-term retention and BYoM terms.

---

👍 VUs Billing Process & Hard Limits

• The platform calculates daily usage per feature.
• VUs are averaged monthly.
• Credits are applied based on the monthly average.
• Increases or decreases in usage are reflected immediately in billing.

---

VUs Credit Price

Hard limits can be configured to control fixed billing exposure.

---

👍 Credit Hard Limit

Hard limits are available per workarea. You can set a VUs hard limit, and the platform will prevent the creation of additional VU-generating items once the limit is reached.

---

Computing Hours @ Arpia Cloud

If your plan is billed using Usage Computing Credits, the following applies:

❗

Computing hours do not apply if your account is billed by Value Units.

---

GPU, Storage, AI LLM & Other Components

---

GPU Droplets

---

LLMs Core Models @ Arpia Cloud

Arpia provides core LLM models from major providers.

• If customers configure their own models (BYM), no token charges apply.
• If no model is provided, ARPIA uses Shared LLM Models billed per Token Consumption.

Token Consumption

Each Shared LLM Model is priced per 1 Million Tokens, the platform's standard Billing Unit. Input and output are metered separately: a request is charged on its own input and output counts, not on a blended rate. Consumption is accumulated across the billing period and charged according to the selected model's published rates. This applies only to Shared LLM Models; BYM configurations are not subject to token charges.

---

📊 Model Classification for Enterprise

Enterprise models are divided into two commercial groupings:

• 🔴 Frontier Models — highest-capability models for demanding reasoning, long-horizon agent work, and complex analysis.
• 🟢 Core Models — fast, economical models for routine enterprise workloads where cost efficiency matters more than peak capability.

⚠️

Both Core and Frontier models are billed Pay-Per-Use by default. Enterprise customers with a contracted agreement should refer to their signed terms, which may include provisions differing from this document.

30 billable models are currently available: 13 Frontier, 17 Core.

---

🔴 ARPIA Frontier LLM Models (Pay-Per-Use)

Always billed separately, per consumption. Not eligible for inclusion under any plan.

---

🟢 ARPIA Core LLM Models (Pay-Per-Use)

Billed per Token Consumption unless otherwise specified in a signed contract.

---

📋 Notes on LLM Pricing

• Billing unit. All rates are per 1,000,000 tokens. Input and output are metered separately; a request is charged on its own input and output counts, not a blended rate.
• Cached input. Prompt-cache reads and writes are counted as input tokens at the model's input rate.
• Context and max output. Context is the total window; max output is the ceiling on a single response. Requests asking for more output than the model allows are clamped to that ceiling.
• Tiers. Frontier and Core are commercial groupings, not provider labels. A model may be reclassified as the market moves without its identifier changing.
• Bring your own key. When a customer supplies their own provider credential, the provider bills them directly and these rates do not apply. The ARPIA governance fee of $1.00 per 1M tokens (input and output combined) applies instead.

---

🎯 Usage Recommendations

1. Use Core Models for 80–90% of operations.
2. Reserve Frontier Models for advanced reasoning.
3. Monitor usage in the dashboard.
4. Configure alerts for both Core and Frontier consumption.

---

📘 BYM (Bring Your Model)

When users configure their own LLM models:

• No token charge applies.
• Only Computing Hours or VUs are billed.
• The ARPIA governance fee of $1.00 per 1M tokens (input and output combined) applies.

---

Storage @ Arpia Cloud

Shared Storage

---

Dedicated Storage

For HA setups, add +1 node per existing node.

---

Credit Billing Policy & Terms

1. Credit Rollover: Annual plans only.
2. No Refunds: Except in case of system error.
3. Invoices are generated on the 5th of each month based on previous month's usage.
4. Payment via Stripe.
5. Prepaid balances are automatically deducted from recurring usage.
6. 2+ overdue invoices may cause service interruption.

---

Account Billing & Usage Support

For assistance, contact billing support via the Support Chat on the ARPIA Web Platform.