Pricing
A plan fee, plus the tokens you use.
Start free in the sandbox. When you subscribe, you pay a monthly plan fee plus input and output tokens at your base model’s rates.
Preview prices
Plan fees and token rates below are placeholders while we benchmark inference costs on real hardware. Final prices will be published before paid plans open, and you’ll see the current price book in your dashboard before you subscribe.
Sandbox
Integrate before you pay.
$0
- Upload and validate adapters
- Create deployments and API keys
- Call the real API endpoints
- Responses are synthetic and labeled — no real inference
- Never billed; daily request limit applies
Developer
Preview priceFor a first client in production.
$19 /month
+ tokens at your model’s rates
- Up to 3 adapters
- Up to 3 active deployments
- Up to 3 clients
- Up to 2 team members
- Real inference on shared capacity
- Per-client reports and CSV export
- No capacity or uptime guarantee during preview
Production
Preview priceFor teams serving many clients.
$79 /month
+ tokens at your model’s rates
- Up to 25 adapters
- Up to 25 active deployments
- Up to 50 clients
- Up to 10 team members
- Real inference on shared capacity
- Per-client reports and CSV export
- No capacity or uptime guarantee during preview
Token rates
Billed per token, shown per million tokens. Token counts come from the inference runtime.
| Base model | Context | Input / 1M tokens | Output / 1M tokens |
|---|---|---|---|
| Qwen3-8BPreview | 32,768 tokens | $0.40 | $1.60 |
Estimate a month
A rough estimate before taxes. Your invoice uses your actual usage.
- Plan fee
- $79.00
- Input tokens
- $8.00
- Output tokens
- $8.00
- Estimated total
- $95.00
Calculated with preview prices.
Billing details
- What counts as a billable token?
- Exactly the prompt and completion token counts the inference runtime reports for each request. Client-supplied counts are ignored. Requests that fail before producing output aren’t billed.
- Can I cap my spend?
- Yes. Owners can set a monthly cap; requests that could exceed it are rejected. How the cap works.
- How do I pay?
- By card through Stripe. We don’t store card details. Invoices are issued monthly.
- Do you offer dedicated capacity?
- Not yet. During the preview all paid plans run on shared capacity.
Integrate today. Pay when it’s real.
The sandbox is free and never billed. Subscribe when you’re ready to serve real traffic.