# OpenAI API: Platform/tokenizer giving very incorrect/inconsistent counts?

**URL:** <https://community.openai.com/t/openai-api-platform-tokenizer-giving-very-incorrect-inconsistent-counts/110235>\
**Category:** API\
**Created:** [March 21, 2023, 12:41am UTC](https://community.openai.com/t/openai-api-platform-tokenizer-giving-very-incorrect-inconsistent-counts/110235 "2023-03-21T00:41:01Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![pmshadow](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/pmshadow/32/146751_2.png) [@pmshadow](https://community.openai.com/u/pmshadow)\
**Post date:** [March 21, 2023, 12:41am UTC](https://community.openai.com/t/openai-api-platform-tokenizer-giving-very-incorrect-inconsistent-counts/110235/1 "2023-03-21T00:41:02Z")

</div>

I was using the token-counter earlier today (about 3h-4h ago, and I felt the results were quite wrong.)  
I had counted the tokens for the same string before, and they were close to 1500. Counted during/after shortage, same string = 2700.  
Counted again right now and they are back to 1500-level.

> **[OpenAI API](https://platform.openai.com/tokenizer)**
>
> An API for accessing new AI models developed by OpenAI

Thanks