Room to explore.
A path to grow.
Account creation is free. Add prepaid credits before real evaluations, and start with a sample as small as 1 item. Connect your company endpoint or use your own OpenRouter or Cloudflare AI Gateway credentials.
Free account setup
Connect an endpoint, explore protocols, and prepare your first sample.
- Custom endpoint integrations
- Completions and decision model support
- Evaluation history and task-level scores
- Scoped API keys
- Versioned benchmark catalog
Free account setup does not include real inference or paid evaluation rows. Fund your wallet before running. Paid requests also need an active monthly platform subscription and a saved card; review the configured rates in your dashboard.
Pro
For teams bringing evaluation into their everyday development workflow.
- Everything in account setup
- Higher evaluation allowances
- More items per evaluation
- Full ready benchmarks within plan limits
- Self-service billing through Stripe
- Manage your subscription in one place
Subscription checkout is available only when billing is configured. Review the monthly plan price in Stripe before paying. A subscription does not add usage credits.
Platform evaluation fees
A platform fee applies to each successfully graded row when you use your company endpoint, OpenRouter key, or Cloudflare AI Gateway credentials.
At a configured rate of $0.001 per row, 1,000 graded rows cost $1 and a 25-row sample costs $0.025. Correct and wrong answers count; failed and unexecuted canceled rows do not. Confirm the configured rate before running.
Your provider account
Your provider bills inference directly. allevals does not add a markup to those provider charges.
Choose an endpoint that supports the benchmark’s required protocol. A provider request may incur cost even if its result cannot be graded; review your provider’s rates before running.
A few things worth knowing.
How do prepaid credits work?
Add credits through a Stripe-hosted payment. Your card is charged for the top-up now; credits appear only after the payment is verified. Evaluation usage is deducted from those credits—not added to a later usage invoice. Reserved credits are unavailable for new work, and new paid requests stop when available credits are exhausted. Your monthly platform subscription is separate.
Can my wallet reload automatically?
Auto-reload is optional and off until you explicitly enable it. Choose a low-credit threshold, a top-up amount, and a maximum automatic reload total per UTC day, then authorize off-session charges to your saved card. You can disable it in billing; replacing the card requires fresh consent. Failed or authentication-required payments add no advance credit. An unresolved automatic payment must be confirmed or reconciled before another manual or automatic charge can begin. You may resume the same existing manual payment or update your card, but updating the card alone does not resolve an uncertain payment.
Who pays for model inference?
With a custom endpoint, your OpenRouter key, or your Cloudflare AI Gateway credentials, your provider bills inference directly. allevals deducts the configured fee for each successfully graded row from your prepaid credits. Start with 5, 10, or 25 items; a full benchmark is never required.
Can I use my own model endpoint?
Yes. Connect a public HTTPS endpoint with an API key and optional custom headers. Each benchmark specifies the endpoint protocol it requires. See the endpoint guide.
Are all catalog benchmarks runnable?
Yes. The catalog only shows benchmarks available on this deployment. Your account still needs a connected endpoint and sufficient credits for paid evaluations.
I already bought a benchmark import package.
Your previous purchase is not a new subscription. Sign in using your purchase email, then use “Restore a previous purchase” on the billing page to securely claim eligible access.