> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ardor.cloud/llms.txt
> Use this file to discover all available pages before exploring further.

# Pricing and Billing Overview

> Ardor uses credits to track Copilot and platform usage. Plans differ by relative included Copilot capacity, rolling usage windows, model access, and extra usage options.

## Credit-Based Pricing Model

Ardor uses credits to make usage easier to understand across Copilot and metered platform resources.

* **Copilot usage** is based on actual model work: input tokens, generated output, cached context, images, and video generation where applicable.
* **Platform resources** are metered by actual infrastructure consumption, such as compute, memory, GPU memory, and storage.

<Info>
  Credits are not a simple count of prompts. A short prompt can consume very little, while a long generation, large context, image, or video task can consume more.
</Info>

### Subscription Plans

Each plan includes a monthly Copilot credit budget and rolling usage windows. The rolling windows are what you see in the Console as 5-hour and weekly usage.

| Plan        | Included Copilot Capacity | Rolling Usage Windows    | Model Access                                           | Extra Usage                       |
| ----------- | ------------------------- | ------------------------ | ------------------------------------------------------ | --------------------------------- |
| **Free**    | Baseline                  | Baseline rolling windows | Gemini 3.1 Pro, Gemini 3.5 Flash                       | Not available                     |
| **Starter** | 4x Free                   | 4x Free rolling windows  | Free models plus GPT-5.4, Grok 4.20, Claude Sonnet 4.6 | Top-ups and extra usage available |
| **Pro**     | 10x Free                  | 10x Free rolling windows | Starter models plus Claude Opus 4.6 and 4.7            | Top-ups and extra usage available |

### How Copilot Burns Credits

Copilot usage is calculated from the finished session events reported by the model provider. The main drivers are:

* **Input tokens** - the prompt, files, context, and instructions sent to the model
* **Output tokens** - generated code, text, reasoning output, and tool instructions
* **Cached input tokens** - reused context, which is usually much cheaper than sending it fresh
* **Media output** - generated images or video seconds, which consume credits faster than text-only work

Generated output usually burns credits faster than input context. Large prompts can also move some models into a higher long-context usage tier.

### Model Credit Impact

This is an approximate guide, not a detailed price sheet. Actual usage depends on prompt size, generated output, context length, caching, and media generation.

| Model                     | Available From | Relative Credit Burn                        | Good Fit                                                           |
| ------------------------- | -------------- | ------------------------------------------- | ------------------------------------------------------------------ |
| **Gemini 3.5 Flash**      | Free           | Low                                         | Frontier agentic coding, multimodal work, and fast iteration       |
| **Gemini 3.1 Pro**        | Free           | Medium; higher on long contexts             | Frontier reasoning, coding, UI work, and long-context synthesis    |
| **GPT-5.4**               | Starter        | Medium to high; higher on long contexts     | OpenAI-style planning, debugging, and code generation              |
| **Grok 4.20**             | Starter        | Low to medium; higher on very long contexts | Marketing copy and exploratory writing                             |
| **Claude Sonnet 4.6**     | Starter        | Medium to high                              | Code review, writing, and structured changes                       |
| **Claude Opus 4.6 / 4.7** | Pro            | Highest                                     | Anthropic frontier reasoning, long-running agents, and code review |

<Note>
  Think of this table as availability and relative burn, not model ranking. Free includes Gemini 3.5 Flash and Gemini 3.1 Pro as frontier-grade models for serious coding, reasoning, multimodal, and long-context work. Opus models are Pro-only because they are significantly more expensive to run.
</Note>

### Extra Usage and Top-Ups

When included Copilot credits or rolling windows are exhausted:

* **Free** workspaces wait for the window to reset or upgrade to a paid plan.
* **Starter** and **Pro** workspaces can use purchased top-up balance when extra usage is enabled and an extra usage limit is configured.
* Extra usage is still capped by your configured limit, so a workspace cannot spend indefinitely by accident.

## Billing Structure

### Subscription vs Pay-as-you-go

Ardor's billing consists of two main components:

* **Subscription plans** - Included Copilot capacity, rolling usage windows, model access, and platform features
* **Pay-as-you-go usage** - Additional usage through purchased balance or metered platform resources where applicable

**How it works:**

1. Your subscription provides monthly included Copilot capacity.
2. Copilot sessions consume credits based on actual model usage.
3. The Console shows active 5-hour and weekly rolling windows.
4. When included usage is exhausted, paid plans can continue through top-ups and configured extra usage.
5. Platform resources are metered separately based on actual resource consumption.

<Note>
  The Console's weekly percentage is the percentage of the current rolling weekly Copilot limit, not a count of prompts. A single large generation can move the percentage more than several small prompts.
</Note>

### Real-Time Cost Monitoring

Ardor provides real-time usage tracking through the dashboard, allowing you to monitor and optimize resource allocation based on demand.

## Example Usage Calculation

**Scenario:** A Starter workspace sends three prompts:

* One short clarification prompt
* One code-generation prompt that reads several files
* One large implementation prompt that produces a lot of output

These do not consume three identical units of usage. The short clarification may consume very little, while the large implementation can consume much more because it sends more context and produces more output.

If the Starter weekly usage shows 33%, that means the workspace has used about one third of its current Starter weekly Copilot window. If the 5-hour window is reached first, Copilot usage pauses until that 5-hour window resets, even if monthly capacity remains. To continue immediately, enable pay-as-you-go extra usage with a configured limit or move to a higher plan with a larger rolling window.

## Payment & Invoicing

Ardor provides transparent billing through our credit-based system:

### Payment Options

* **Monthly subscriptions** - Fixed plan fees with included Copilot capacity
* **Additional credit purchases** - Top-up credits when needed (paid plans only)
* **Automatic billing** - Credits are deducted in real-time as you use services

### Invoice Details

* **Plan subscription fees** - Charged monthly for included credits and platform features
* **Additional credit purchases** - One-time charges when you buy extra credits
* **Usage transparency** - Detailed breakdown of credit consumption by service type
* **Real-time tracking** - Monitor credit usage and remaining balances in your dashboard

<Info>
  Usage charges are converted to credit deductions, making it easier to understand and predict usage across Ardor services.
</Info>