Rate limits

#Generation rate limits

Rate limits for model inference requests depend on your account tier:

Free-tier teams granted free credits by an operator are raised to 200 RPM, and keep that limit until an operator removes it; it does not expire when the credits run out. The pay-as-you-go credit-purchase floor is 250 RPM, so a real purchase always buys more headroom than a free grant. Per-team overrides for custom teams are configured by an operator; contact us to request one.

#Batch API rate limits

Batch processing rate limits are separate from generation rate limits. See the Batch API for details.

#Ad-supported model spend limits

Ad-supported models (for example moonshotai/kimi-k3.1-pro-with-ads) have two spend limits on top of the generation rate limit:

• an hourly limit shared by all users of ad-supported models;
• a daily limit for each user (or each team, for requests without a user).

Both reset at the start of the UTC hour or day. Once either is reached, requests to ad-supported models return HTTP 429 with error type GravityAdsDiscountCap, a message naming the limit and when it resets, and a Retry-After header. Standard models are not affected, so you can switch to the standard model instead.

#Payload size limits

• Request body: up to 32 MiB (33,554,432 bytes), including inline images. Larger requests are rejected with HTTP 413.
• Non-streaming response body: up to 32 MiB. A provider response larger than that is refused with HTTP 502 and error type UpstreamResponseTooLarge. Use stream: true, or smaller batches for embeddings, to stay under it.

#When you hit a limit

You receive a 429 Too Many Requests response. Back off and retry with exponential backoff.

#Increase your limits

If you need higher rate limits, contact support, use the support chat, or talk to the team about a custom tier.