scenarios/workload-genai/policies/fragments/rate-limiting/rate-limiting-by-tokens.xml

<fragment>
    <!-- Rate limit configuration at a subscription level. Works for Azure OpenAI endpoint, 
         both streaming and non-streaming endpoints.
         allows 500 tokens per 60 seconds. -->
    <azure-openai-token-limit counter-key="@(String.Concat(context.Subscription.Id,"-max-token"))"
        tokens-per-minute="500"
        estimate-prompt-tokens="true"
        remaining-tokens-header-name="x-apim-max-remaining-tokens"
        tokens-consumed-header-name="x-apim-max-consumed-tokens"/>
</fragment>