# Performance of hosted inference API

**URL:** <https://discuss.huggingface.co/t/performance-of-hosted-inference-api/3657>\
**Category:** Beginners\
**Created:** [February 16, 2021, 6:21am UTC](https://discuss.huggingface.co/t/performance-of-hosted-inference-api/3657 "2021-02-16T06:21:25Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![bala1802](https://avatars.discourse-cdn.com/v4/letter/b/8797f3/32.png) [@bala1802](https://discuss.huggingface.co/u/bala1802)\
**Post date:** [February 16, 2021, 6:21am UTC](https://discuss.huggingface.co/t/performance-of-hosted-inference-api/3657/1 "2021-02-16T06:21:25Z")

</div>

For any one fine-tuned GPT-2 model, would like to understand the performance (response time and number of concurrent requests that can be served) of the hosted inference API of Hugging Face in ‘Lab’ mode and ‘Startup’ mode?