Cost Tracking
The Web UI tracks token usage and estimated costs for API-backed models.
Pricing Database
Saturn includes pricing data for 20+ models covering OpenAI, Anthropic, Google, and common open-weight models served through paid APIs. Prices are stored as cost per million tokens (input and output separately).
Local models (Ollama, LM Studio) report zero cost.
Per-Message Cost
After each response, a small cost badge appears below the message showing:
- Input tokens used
- Output tokens used
- Estimated cost (e.g.,
$0.003)
Token counts are read from the API response headers or body when available.
Daily Spend
The settings overlay shows a Budget section with:
| Metric | Description |
|---|---|
| Today's spend | Sum of all message costs since midnight (local time) |
| Total spend | Cumulative across all sessions |
| Daily limit | Configurable cap (default: unlimited) |
Both daily and total spend are stored in localStorage.
Spend Limit
Set a daily spend limit in the settings overlay. When the limit is reached:
- A warning banner appears at the top of the chat
- New messages are blocked until the next day or the limit is raised
- The send button is disabled with a tooltip explaining why
Tip
Set a limit before using expensive models (GPT-4o, Claude Opus) to avoid surprise costs on metered API keys.
Budget Summary
The budget summary in settings shows a breakdown:
- Per-model spend for the current day
- A bar chart of daily spend over the last 7 days
- The configured limit with a progress indicator