Service Level Agreement

Effective Date: November 5, 2026

This Service Level Agreement ("SLA") is part of the Terms of Service between you and ModelRiver Inc. ("ModelRiver," "we," "us"). It applies to organizations on a Scale or Custom plan. Words defined in the Terms mean the same here.

Who it applies to

If your Custom order form sets a different uptime commitment or different credits, the order form wins.

What's covered

The SLA covers these services, as listed on our status page at status.modelriver.com:

• API: inference and OpenAI-compatible endpoints.
• Async requests and webhooks: background jobs, streaming and webhook delivery.

The web console (console.modelriver.com) isn't covered. Your apps keep working through the API when the console is unavailable.

How uptime is measured

• Our monitors check each covered service once a minute.

• A minute counts as downtime when the check for that service fails: it returns an error, or doesn't respond within 10 seconds.

• Monthly uptime for a service is the share of minutes in the calendar month (UTC) that weren't downtime:

  Monthly uptime % = (minutes in the month − downtime minutes) ÷ minutes in the month × 100

• If the API and async services both miss the commitment in the same month, the credit is based on whichever had the lower uptime. Credits for one month are never added together across services.

The uptime history on our status page is the record we use. If you saw downtime our monitors missed, send us your evidence (times, request IDs, error responses) with your claim and we'll review it.

Service credits

If monthly uptime for a covered service falls below 99.9%, you can claim a credit of a share of that month's plan fee:

• Plan fee means the fee for your plan for that month, not including overage charges or taxes.
• Credits are applied to your next invoice. They aren't paid out in cash and can't be transferred.
• The credits for any one month are capped at 100% of that month's plan fee.
• For a plan billed yearly, the monthly plan fee is the yearly fee divided by 12.
• Service credits are your only remedy for downtime of the covered services.

How to claim a credit

Email [email protected] from an owner's address within 30 days after the end of the affected month, with:

• your organization name,
• the dates and times (with time zone) you were affected, and
• any request IDs or error responses you have.

We'll reply within 10 business days. Approved credits appear on your next invoice. Your account must be paid up to date to receive a credit.

What doesn't count as downtime

Downtime doesn't include time a covered service is unavailable or failing because of:

• Scheduled maintenance we announce on our status page at least 48 hours in advance.
• AI providers: errors, timeouts or outages from OpenAI, Anthropic, Google or any other provider you send requests to. Set up fallbacks in your workflows so another provider takes over.
• Your systems: your own servers, network, code or configuration, including webhook or callback URLs that don't respond or return errors.
• Limits working as intended: rate limits, spend limits, project quotas, plan limits or guardrails blocking a request.
• Your account's status: suspension or restriction under the Terms, such as for unpaid invoices or misuse.
• Test mode, the Playground and features marked beta or preview.
• Events outside our reasonable control, such as natural disasters, war, government action, or failures of the internet outside our and our subprocessors' networks.

Changes to this SLA

We may update this SLA. We'll post changes on this page at least 30 days before they take effect. A change that lowers the uptime commitment or the credits doesn't apply until your next renewal, and never during a Custom contract's term.

Contact

Questions about this SLA: [email protected]

Last updated: November 5, 2026