Merge pull request #97157 from swmachan/master

Clarity on latency

Author: PRMerger9

articles/cognitive-services/Translator/request-limits.md

@@ -8,7 +8,7 @@ manager: nitinme
 ms.service: cognitive-services
 ms.subservice: translator-text
 ms.topic: conceptual
-ms.date: 06/04/2019
+ms.date: 11/25/2019
 ms.author: swmachan
 ---
 
@@ -53,7 +53,7 @@ These limits are restricted to Microsoft's standard translation models. Custom t
 
 ## Latency
 
-The Translator Text API has a maximum latency of 15 seconds using standard models. Translation using custom models has a maximum latency of 25 seconds. By this time you'll have received a result or a timeout response. Typically, responses are returned in 150 milliseconds to 300 milliseconds. Response times will vary based on the size of the request and language pair. If you don’t receive a translation or an [error response](https://docs.microsoft.com/azure/cognitive-services/translator/reference/v3-0-reference#errors) within that timeframe, you should check your network connection and retry.
+The Translator Text API has a maximum latency of 15 seconds using standard models and 120 seconds when using custom models. Typically, responses *for text within 100 characters* are returned in 150 milliseconds to 300 milliseconds. The custom translator models have similar latency characteristics on sustained request rate and may have a higher latency when your request rate is intermittent. Response times will vary based on the size of the request and language pair. If you don’t receive a translation or an [error response](https://docs.microsoft.com/azure/cognitive-services/translator/reference/v3-0-reference#errors) within that timeframe, please check your code, your network connection and retry. 
 
 ## Sentence length limits