Max Tokens: k
Set your own OpenAI API Key, so that your account is charged for the use.
If you don't have an API Key, register to OpenAI
Platform and create one there.
You can optionally set the API Key via the apikey URL argument and save as a bookmark or to the desktop,
but be careful not disclosing it to others.
The URL arguments darkmode and markdown can be used for respective setting. Set to '1' or 'true' to
Hover over settings and buttons for a more thorough description.
GPT-3.5 Turbo: More likely to follow just one part of a complex multi-part instruction. GPT-3.5 Turbo
returns outputs with lower latency and costs much less per token.
GPT-4: Generally performs better on a wide range of evaluations. In particular, it's more capable at
carefully following complex instructions. It's also less likely to make up information, a behavior known
The following value indicates how many tokens a complete conversation can have, including previous
prompts and responses.
User: Provides requests or comments for the assistant to respond to.
System: Sets the behavior of the assistant. For example, you can modify the personality of the assistant
or provide specific instructions about how it should behave throughout the conversation. If used, write
this first in a new conversation.
Assistant: Used by the GPT responses, but can also be written by you to give examples of desired
Saved or externally created JSON and CSV files can be loaded to continue a conversation.
Only JSON will include model settings, so preferably use that format for archiving and continuation of
Text, HTML and RTF are intended for use by other applications and can't be loaded into Guppy.
Write your prompts as tasks rather than as questions.
Be as descriptive and clear as possible and spell correctly to avoid confusion.
You can continue conversations knowing that GPT is aware of the previous prompts and responses.
Nothing is stored centrally about API Key, prompts or responses, or anything else for that matter.
We only collect anonymous usage statistics at loading of the page.