API Rate Limits

​

Overview

Rate limits are applied per member, per gateway pair, and per quoting type (order entry or mass quoting). Each gateway within a pair enforces its limits independently; violating a rate limit on gateway A does not affect gateway B. Orders and mass quotes consume tokens from separate buckets, allowing you to manage order and quoting activity independently. Cancellations are exempt and never consume tokens.

​

Leaky Bucket Algorithm

Starbase uses a leaky bucket algorithm. Each bucket has two parameters: Every request adds tokens to the bucket. When the bucket is full, subsequent requests are rejected until enough tokens have drained. The bucket drains at a constant rate of R tokens per millisecond. This translates to two intuitive limits:

1. Burst — The maximum number of messages you can send instantaneously: B / (tokens per request)
2. Steady state — The maximum sustained throughput: R / (tokens per request) × 1000 messages per second

​

Token Cost per Request

Each request type consumes a fixed number of tokens:

​

Default Rate Limits

The table below shows the default rate limits. These defaults apply to all members unless overridden. See Underlying Tiers for the full tier classification. All five product groups share the same defaults:

​

Per-Member Overrides

Rate limits can be increased on a per-member basis at the discretion of Deribit, based on current or expected contribution to liquidity. Overrides are configured per product group and quoting type independently. The following tables illustrate an example override: Bucket parameters — R in tokens/ms, B in max tokens Rate limits — burst in messages, steady in messages/s Rate limits are assigned based on the technical needs of the strategy. In practice, the majority of Starbase’s total capacity is allocated to market makers who provide continuous two-way liquidity across many instruments. The spread of the instruments quoted influences the allocation: quoting a perpetual future at a bid-ask spread of less than 1 basis point will receive a much larger rate limit allocation than a dated future trading at a spread of 10 basis points.

​

Rate Limit Violations

When a bucket is full, new orders, amends, and mass quotes are rejected. The reject response indicates the rate limit was exceeded. Cancellations are never rejected due to rate limits.

​

Gateway Redundancy

Each gateway pair consists of two independent gateways (A and B). Both gateways enforce identical and independent rate limits. This design encourages clients to load balance across both gateways without requiring the gateways to synchronize rate limit state.

​

Review Schedule

Rate limit allocations are reviewed periodically. The contribution of trading members is assessed to ensure their performance aligns with allocated rate limits. Rate limits are not adjusted automatically and only change with prior notice.

​

Other Limits

​

API Key Limits

Each member has a default limit of 10 API keys, allowing 10 connections per gateway. This limit can be increased at the discretion of Deribit. Please reach out to Support if this limit is too low for your activities.

​

Open Order Limits

The number of orders and quotes active at any time is limited per member. The default is 2,000 and can be increased at the discretion of Deribit. Please reach out to Support if this limit is too low for your activities.