GPT for Sheets reference

Discover all available functions in GPT for Sheets, along with simple examples and an extensive list of paramaters. If you are looking for usage examples, click here.

GPT function

The simplest function to start using GPT in Sheets. Outputs the result in a single cell.

How to use

Syntax	=GPT(prompt, [value], [temperature], [max_tokens], [model])
Copy/paste example	=GPT("Write a tagline for an ice cream shop.")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	A response to your prompt in a single cell

Parameters

Parameter	Definition
prompt (cannot be empty)	Text, cell or range containing the prompt. Examples: "Write a tagline for an ice cream shop." A1 A1:C3
(optional) value	Text, cell or range you want your prompt to apply to
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

See more usage examples.

GPT_LIST / GPT_HLIST function

Like GPT, but outputs the results in a column. Very practical when the output is a list.

• Use GPT_LIST to return options listed vertically.
• Use GPT_HLIST to return options listed horizontally.

How to use

Syntax	=GPT_LIST(prompt, [value], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_LIST("Give me 5 good short ads about spreadsheets.")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	As many responses to your prompt as you asked, one response per cell

Parameters

Parameter	Definition
prompt (cannot be empty)	Text, cell or range containing the prompt. Examples: "Write a tagline for an ice cream shop." A1 A1:C3
(optional) value	Text, cell or range you want your prompt to apply to
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

See more usage examples.

GPT_SPLIT / GPT_HSPLIT function

Splits text semantically, such as by section, paragraph, sentence, customer…

• Use GPT_SPLIT to return outputs listed vertically.
• Use GPT_HSPLIT to return outputs listed horizontally.

How to use

Syntax	=GPT_SPLIT(text, split_by, [temperature], [max_tokens], [model])
Copy/paste example	=GPT_SPLIT("Hello! How are you? Have you heard of GPT for Sheets?", "sentences")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	Input text is split as requested, one element per cell

Parameters

Parameter	Definition
text (cannot be empty)	Input text to split
split_by (cannot be empty)	How to split the text. Examples: sections paragraphs sentences punctuation
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_FILL function

Magically fill or clean a range from a few examples.

How to use

Syntax	=GPT_FILL(examples, [inputs], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_FILL(S4:U5,S6:S14)
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	The empty part of the range will be filled based on the examples provided

Parameters

Parameter	Definition
examples	Range containing complete examples that GPT should learn from
(optional) inputs	Range containing incomplete data that GPT should complete from the examples
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

See more usage examples.

GPT_TABLE function

Like GPT, but outputs the results in a table. Very practical when the output is a table.

How to use

Syntax	=GPT_TABLE(prompt, [head], [inputs], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_TABLE("top 10 most eaten fruits and their nutrition data")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	The response to your prompt in the shape of a table

Parameters

Parameter	Definition
prompt (cannot be empty)	Text, cell or range containing the prompt. Examples: "top 10 most eaten fruits and their nutrition data" A1 A1:C3
(optional) head	Range containing at least one row. the first row should be headers, the following rows can be used to specify examples If left empty, headers will be generated automatically.
(optional) inputs	Range containing partial rows to complete with GPT_TABLE
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_MAP function

Map the values of two columns by similarity.

How to use

Syntax	=GPT_MAP(search_keys, range, [confidence], [stats], [top_k])
Copy/paste example	=GPT_MAP(A1:A10,B1:B4)
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	A column of the best matches to each value in the range

Parameters

Parameter	Definition
search_keys (cannot be empty)	The values that you want to look for in the range
range (cannot be empty)	The range in which to look for values that are similar to search keys
confidence (optional)	Set a similarity threshold: only values with a superior similarity score are mapped. Must be comprised between 0 and 1
stats (optional)	"true" to display similarity scores
top_k (optional)	Number of matching values to return per search_key (default: 1 - only the best match)
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_FORMAT function

Format dates, currencies, addresses, names, etc. Fix capitalization. And so much more.

How to use

Syntax	=GPT_FORMAT(input, target_format, [source_format], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_FORMAT("14 February 71", "iso")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	The converted input in a single cell unless it is converted a table, which outputs a table

Parameters

Parameter	Definition
input (cannot be empty)	Text, cell or range containing the input to be converted. Examples: "14 February 1971" A1 A1:C3
target_format (cannot be empty)	Target format for the input. Examples: "iso", "currency iso", "international phone number", "lowercase/uppercase/camelcase/snake case/title case/sentence case"…
(optional) source_format	Source format of the input. Optional, but including it can yield better results.
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_EDIT function

Applies the given task to the given text. The default ask is to fix grammar and spelling.

How to use

Syntax	=GPT_EDIT(text, [task], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_EDIT("For sum reezon thoose nunsberz arnt addin up")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	The edited text in a single cell

Parameters

Parameter	Definition
text (cannot be empty)	Text, cell or range to be edited. Examples: "For sum reezon thoose nunsberz arnt addin up" A1 A1:C3
(optional) task	Text, cell or range specifying how the text should be edited. Defaults to fixing grammar and spelling. Other examples: "Make it funnier" "Make it shorter" "Make it sound formal" "Add an ending"
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_TAG function

Applies user-defined tags to a given text.

How to use

Syntax	=GPT_TAG(text,[tags], [top_k], [examples], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_TAG("I love chocolate", "food, positive, negative")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	The applicable tags as csv in a single cell

Parameters

Parameter	Definition
text (cannot be empty)	Text, cell or range. Examples: "I love chocolate" A1 A1:C3
(optional) tags	Comma-separated tags or range of tags applicable. If left empty, GPT_TAG will automatically come up with the tags
(optional) top_k	Maximum number of tags to return
(optional) examples	A table of examples in two columns. The first column should contain the inputs and the second column should contain the outputs.
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_CLASSIFY function

Classifies a given text into a single category.

This function can use the text-embedding-ada-002 model for stricter application of categories and cheaper cost.

How to use

Syntax	=GPT_CLASSIFY(text, categories, [examples], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_CLASSIFY("banana", "fruit, vegetable")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	The best applicable category

Parameters

Parameter	Definition
text (cannot be empty)	Text, cell or range. Examples: "banana" A1 A1:C3
categories (cannot be empty)	Comma-separated categories or range of categories to choose from. Only the most relevant category will be returned.
(optional) examples	A table of examples in two columns. The first column should contain the inputs and the second column should contain the outputs.
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_EXTRACT function

Extracts data (like email addresses or company names) from a text. Outputs as comma-separated values.

How to use

Syntax	=GPT_EXTRACT(text, to_extract, [temperature], [max_tokens], [model])
Copy/paste example	=GPT_EXTRACT("I worked 12 years at Amazon.com at Apple", "companies")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	The extracted entities as comma-separated values in a single cell

Parameters

Parameter	Definition
text (cannot be empty)	Text, cell or range to extract data from. Examples: "I worked 5 years at Amazon.com and then 3 years at Apple" A1 A1:C3
to_extract (cannot be empty)	Text, cell or horizontal range specifying what you want to extract. Examples: "email addresses" B3:E3
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_SUMMARIZE function

Summarizes text according to a given format (three sentences by default).

How to use

Syntax	=GPT_SUMMARIZE(text, [format], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_SUMMARIZE("some long content here")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	A summary in the requested format in a single cell

Parameters

Parameter	Definition
text (cannot be empty)	Text, cell or range to summarize. Examples: "Some long content" A1 A1:C3
(optional) format	Format of the summary. Examples: "key takeaways" "action items" "two paragraphs" "100 words" "tweets"
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_TRANSLATE function

Translates text from one language into another. Works better if you specify the source language.

How to use

Syntax	=GPT_TRANSLATE(text, target_language, [source_language], [instructions],[temperature], [maxTokens], [model])
Copy/paste example	=GPT_TRANSLATE("Cool off with our delicious treats!", "spanish")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	The requested translation in a single cell

Parameters

Parameter	Definition
text (cannot be empty)	Text, cell or range to translate. Examples: "Cool off with our delicious treats!" A1 A1:C3
target_language (cannot be empty)	Language of the output. Examples: "spanish", "chinese", "french"
(optional) source_language	Language of the input. Example: "english"
(optional) instructions	Additional instructions for the translation. Can include a dictionary of translations.
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_CONVERT function

Converts from one format to another. For instance json to table, or xml to json.

How to use

Syntax	=GPT_CONVERT(input, target_format, [source_format], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_CONVERT("<ul><li><p>Name: Joe</p><p>Age: 27</p></li><li><p>Name: Jane</p><p>Age: 28</p></li></ul>", "json")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	The converted input in a single cell unless it is converted a table, which outputs a table

Parameters

Parameter	Definition
input (cannot be empty)	Text, cell or range containing the input to be converted. Examples: "<ul><li><p>Name: Joe</p><p>Age: 27</p></li><li><p>Name: Jane</p><p>Age: 28</p></li></ul>" A1 A1:C3
target_format (cannot be empty)	Target format for the input. Examples: "json", "html", "markdown", "json-ld", "table", "csv"
(optional) source_format	Source format of the input. Optional, but including it can yield better results. Examples: "json", "html", "markdown", "json-ld", "table", "csv"
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_CODE function

Generates code to accomplish the given task. Output is a function.

How to use

Syntax	=GPT_CODE(task, [language], [temperature], [max_tokens], [model])
Copy/paste example	=GPT_CODE("concatenate two strings", "Javascript")
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	Code that implements task in a single cell

Parameters

Parameter	Definition
task (cannot be empty)	Text, cell or range specifying what the generated code should do. Examples: "check that a number n is odd" "compute a sha256 hash" "bubble sort"
(optional) language	Programming language the code needs to be written in.
(optional) temperature, max_tokens, model	temperature is a number between 0 and 1. Use it to control creativity. max_tokens is a number between 0 and model limit. Decrease if your request times out. Increase if your response is cut. Go to temperature, max_tokens and model to learn more.

GPT_CREATE_PROMPT helper function

Like CONCATENATE, but inserts spaces and newlines automatically. Helps you create a prompt from scattered, disjointed cells and ranges. You can then use that prompt in GPT functions. This function doesn’t call GPT and thus doesn’t incur any cost.

How to use

Syntax	=GPT_CREATE_PROMPT(arg1, [arg2], [arg3], [...argn])
Copy/paste example	=GPT_CREATE_PROMPT(A1:B1, A2, C3:D5, E1:E6)
⚠️ Localization	In some locales, you might need to use ";" to separate parameters instead of ","
Output	A prompt ready to use in GPT, GPT_LIST or GPT_TABLE

Parameters

Parameter	Definition
arg1 (cannot be empty)	Cell or range to concatenate A1 A1:C3
(optional) arg2 - argn	Cell or range to concatenate A2 G2:H4

temperature, max_tokens and model

These parameters are the same for all GPT functions. They are always optional.

Parameter	Definition
(optional) temperature	Number between 0 and 1 that governs the creativity of GPT: 0 makes GPT strictly follow the prompt (default) 0.5 makes GPT slightly creative 1 makes GPT very creative Note: In some locales, you might need to use "0,5" instead of "0.5"
(optional) max_tokens	The maximum response size in tokens. Must be a round number greater than 0 and lower than the model limit. Defaults to 4096 - prompt tokens for gpt-3.5-turbo Defaults to 500 for other base models to prevent time out risk Learn more about tokens here.
(optional) model	All text OpenAI base and fine-tuned models are available by their name. The most popular base models are the following: gpt-3.5-turbo: (default) Most capable GPT-3.5 model that backs the official ChatGPT interface. Can do any task the other models can do, faster and with higher quality. 10x cheaper than text-davinci-003. gpt-4 or gpt-4-32k: (if you have access) Latest and most powerful OpenAI model with broader general knowledge than other models and advanced reasoning capabilities. Up to 60x more expensive and much slower than gpt-3.5-turbo. text-davinci-003: Very capable GPT-3 model. Can do any task the other models can do, often with higher quality, longer output and better instruction-following. Also supports inserting completions within text. text-curie-001: Less capable, but faster and lower cost than Davinci. text-babbage-001: Less capable, but faster and lower cost than Curie. text-ada-001: Least capable, but fastest and lowest cost of all. Note: GPT_CLASSIFY can use text-embedding-ada-002