Introduction

To look at the limit and which model are exclusive to the advanced tier: https://www.notion.so/AI-Model-Rate-Limits-With-Advanced-Add-on-1cd8332d3ff380169b3adb2754ab23c0?pvs=4#1cd8332d3ff3801b9d9eed75e0541b5d
This comparison is updated personally. It may not be up to date
- For a complete and updated comparison with different benchmark visit this link

Context

They represent how many tokens the ai model support
Raycast ai website and built-in comparison tool only show the input context window
Each token is about 3/4 of an English character. This is why when using raycast and comparing the model you can see an estimation of “pages”
Most models analyze the token as you tipe and not when submitting, this exponentially speed the response time. The more tokens the input given contains the bigger the speed benefit is
They are divided into input and output tokens
- Input: How much text you can ask the model
- Output: How much text the ai can give back to you
What to consider
- Most models have an input token larger than the output token.
- If you find that the response is not complete is mostly because the input you gave is larger than the output it can provide. For example you gave a code snippet that respect the input token window, expected to receive the same code back but it does not happen because your entire code is more than the supported output tokens