Salesforce Einstein · Rate Limits

Salesforce Einstein Rate Limits

Salesforce Einstein consumption is limited at two layers. Calls into the core platform APIs inherit the per-org 24-hour API request allowance scaled by edition + license count. Generative Einstein features (Copilot, Prompt Builder, Trust Layer) additionally meter against per-org / per-user generative request budgets defined per add-on SKU. Exact published numbers were not retrievable in this run; this artifact captures the model and points to Salesforce's published limits.

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

It captures 3 rate-limit definitions, measuring requests_per_day, requests_per_window, and predictions_per_month.

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

Tagged areas include CRM, AI, Salesforce, and Rate Limiting.

3 Limits Throttle: 429
CRMAISalesforceRate Limiting

Limits

Per-org 24-hour API request allowance org
requests_per_day
see Salesforce App Limits Cheatsheet (varies by edition + license count)
Einstein API calls into the core platform consume the same 24-hour bucket as REST / Bulk / Tooling.
Einstein generative request quota org
requests_per_window
see Einstein add-on documentation (per-org / per-user budgets)
Copilot, Prompt Builder, and Trust Layer features have additional generative request budgets defined per add-on SKU.
Einstein Vision / Language predictions account
predictions_per_month
see Einstein Platform Services documentation
Legacy Einstein Vision / Language APIs document monthly prediction allowances per plan.

Policies

Edition + license scaling
Core API allowance scales with edition and license count.
Add-on quotas
Generative Einstein features add their own quotas on top of the core allowance; exhaustion affects only generative endpoints.
Trust Layer governance
The Einstein Trust Layer enforces masking, audit, and zero-retention policies that affect throughput characteristics of generative calls.

Sources