Braintrust · Rate Limits

Braintrust Data Rate Limits

Braintrust documents system-wide operational limits rather than a single global requests-per-second figure for the REST API. Published limits cover function execution throughput, query and code-execution timeouts, and a per-span payload size cap. Usage allotments (processed data, scores) and data retention are governed by the plan tier. Enterprise plans can raise these limits via custom agreements.

Braintrust Data Rate Limits is the machine-readable rate-limit profile for Braintrust on the APIs.io network, conforming to the API Commons Rate Limits specification.

It captures 7 rate-limit definitions, measuring operations, seconds, megabytes, gigabytes, and scores.

The profile also includes 3 backoff/retry policies defined and response codes documented for throttled.

Tagged areas include AI, LLM, Evaluation, Observability, and LLMOps.

7 Limits Throttle: 429
AILLMEvaluationObservabilityLLMOpsRate LimitingQuotasThrottling

Limits

Function Executions account
operations
10000 per 10 seconds
System-wide cap on function execution throughput.
Query Timeout request
seconds
30
Maximum duration for a query before it times out.
Inline Code Timeout request
seconds
240
Maximum execution time for inline (4-minute) code functions.
Bundled Code Timeout request
seconds
30
Default maximum execution time for bundled code functions.
Per-Span Payload request
megabytes
20
Maximum size of an individual logged span / event payload.
Processed Data Allotment account
gigabytes
1 (Starter) / 5 (Pro) / custom (Enterprise) per month
Monthly included processed-data volume; overage billed per GB.
Scores Allotment account
scores
10k (Starter) / 50k (Pro) / custom (Enterprise) per month
Monthly included scores; overage billed per 1k.

Policies

Tiered Limits
Allotments and retention raise from Starter to Pro to Enterprise; Enterprise supports custom limits.
Backoff Strategy
Clients should implement exponential backoff with jitter on HTTP 429 and honor any Retry-After header.
Payload Sizing
Keep individual span payloads under 20 MB; split or summarize large inputs/outputs before logging.

Sources