- Maximum requests per day
- 20,000
- Concurrency at full minute rate
- 10
Capacity assumes steady traffic and continuous availability. It is not a traffic forecast or a recommended safety margin.
How do RPM and RPD translate to daily capacity?
Multiply requests per minute (RPM) by 1,440 minutes per day. If a requests-per-day (RPD) limit also applies, the smaller value is the daily ceiling. Leave RPD blank to estimate capacity from the minute limit alone; a value of zero means no requests are allowed.
How much concurrency does that require?
At a steady minute rate, average concurrent requests equal RPM multiplied by average request duration in seconds, divided by 60. The calculator rounds up to a whole request. Streaming duration includes the time until the response finishes, not just the time to its first token.
Concurrency here describes the full minute rate, even when a daily cap prevents sustaining it all day. Real services need headroom for bursts, slow requests and retries. Token limits and hardware throughput may impose lower ceilings.
Should the limit be global or per organization?
A global limit covers the combined traffic sent to your deployment. A per-organization limit gives each organization its own allowance, so total traffic can grow with the number of organizations. Choose the scope your infrastructure and commercial terms support. Platform limits still apply.
Does a capacity estimate predict traffic?
No. Capacity is a ceiling under the stated assumptions. The public model rankings help you explore usage on LLM Gateway; your own routed traffic appears in Airside. Follow the listing guide to configure deployment limits.