Model & usage pricing
AI usage is billed per token and metered monthly. You pay each provider's published list price plus a small platform fee. There is no fixed monthly number — your total depends on which models you run and how much you use. Pick the model per task: a cheaper, faster model for routine PRs, a frontier model for the hard ones.
Language models
USD per 1,000,000 tokens. Rates shown include a small platform fee.
| Model | Input | Cache write | Cache read | Output |
|---|---|---|---|---|
| Claude Opus 4.8 | $5.50 | $6.88 | $0.55 | $27.50 |
| Claude Opus 4.7 | $5.50 | $6.88 | $0.55 | $27.50 |
| Claude Opus 4.6 | $5.50 | $6.88 | $0.55 | $27.50 |
| Claude Opus 4.5 | $5.50 | $6.88 | $0.55 | $27.50 |
| Claude Sonnet 5 | $2.20 | $2.75 | $0.22 | $11.00 |
| Claude Sonnet 4.6 | $3.30 | $4.13 | $0.33 | $16.50 |
| Claude Sonnet 4.5 | $3.30 | $4.13 | $0.33 | $16.50 |
| Claude Haiku 4.5 | $1.10 | $1.38 | $0.11 | $5.50 |
| Claude Opus 4.1Legacy | $16.50 | $20.63 | $1.65 | $82.50 |
| Claude Opus 4Legacy | $16.50 | $20.63 | $1.65 | $82.50 |
Prompt-cache writes shown at the 5-minute rate (1.25× input). A 1-hour write is 2× input; cache reads are 0.1× input. Caching makes repeat context dramatically cheaper — we use it aggressively so you pay cache-read rates on the parts of your codebase that do not change between reviews.
Budget controls
Usage-based pricing is variable, so the app gives you controls to cap it. Set a monthly limit per workspace and the platform enforces it.
- •Real-time usage tracking — per model, per workspace, per period.
- •Spend alerts at thresholds you set.
- •Hard spend limits that stop usage before it crosses your ceiling.
Model rates from Anthropic API pricing (platform.claude.com/docs/en/about-claude/pricing), verified 2026-09-12. You pay provider list price plus a small platform fee. Provider prices change; we update this page when they do.