# Use images in GPT function prompts in Sheets Use images in your prompts with [GPT for Sheets – Autosheet](https://autosheet.com/docs/gpt-for-sheets) and [vision models](https://autosheet.com/docs/resources/guides/concepts#vision-model). These models are multimodal and can process both text and images as input. > **Tip** > > Don't have image URLs yet? Learn how to [host images on Google Drive for AI processing](https://autosheet.com/docs/gpt-for-sheets/gpt-functions/use-images-in-prompts#host-images-on-google-drive-for-ai-processing). > **Info** > > Supported image formats: PNG, JPEG, WebP, non-animated GIF ## Supported vision models We recommend using **gpt-5.4** to analyze images, but you can use any vision model supported by Autosheet. **Click here for a list of all vision models supported by GPT for Sheets – Autosheet** | Provider | Model | Built-in | Provider-connected | | --------- | --------------------------- | -------- | ------------------ | | OpenAI | gpt-4.1 | Yes | Yes | | OpenAI | gpt-4.1-mini | Yes | Yes | | OpenAI | gpt-4o | Yes | Yes | | OpenAI | gpt-5 | Yes | Yes | | OpenAI | gpt-5-mini | Yes | Yes | | OpenAI | gpt-5-nano | Yes | Yes | | OpenAI | gpt-5.1 | Yes | Yes | | OpenAI | gpt-5.2 | Yes | Yes | | OpenAI | gpt-5.4 | Yes | Yes | | OpenAI | gpt-5.4-mini | Yes | Yes | | OpenAI | gpt-5.4-nano | Yes | Yes | | OpenAI | gpt-5.5 | Yes | Yes | | OpenAI | gpt-5.6-luna | Yes | Yes | | OpenAI | gpt-5.6-sol | Yes | Yes | | OpenAI | gpt-5.6-terra | Yes | Yes | | OpenAI | gpt-6-astra | No | Yes | | OpenAI | gpt-6-luna | Yes | Yes | | OpenAI | gpt-6.1-sol | Yes | Yes | | OpenAI | gpt-6.1-terra | Yes | Yes | | OpenAI | o3 | Yes | Yes | | Anthropic | claude-4.1-opus | No | Yes | | Anthropic | claude-4.5-haiku | Yes | Yes | | Anthropic | claude-4.5-opus | Yes | Yes | | Anthropic | claude-4.5-sonnet | Yes | Yes | | Anthropic | claude-4.6-opus | Yes | Yes | | Anthropic | claude-4.6-sonnet | Yes | Yes | | Anthropic | claude-4.7-opus | Yes | Yes | | Anthropic | claude-4.8-opus | Yes | Yes | | Anthropic | claude-5-opus | Yes | Yes | | Anthropic | claude-5-sonnet | Yes | Yes | | Anthropic | claude-5.5-haiku | Yes | Yes | | Anthropic | claude-5.5-opus | Yes | Yes | | Anthropic | claude-5.5-sonnet | Yes | Yes | | Anthropic | claude-fable-5 | No | Yes | | Anthropic | claude-fable-5.1 | No | Yes | | Mistral | mistral-medium | No | Yes | | xAI | grok-4-1-fast | No | Yes | | xAI | grok-4-1-fast-non-reasoning | No | Yes | ## Prompt images Use the **GPT_VISION** function to describe or analyze images in individual cells. ### Using the model switcher 1. [Open GPT for Sheets – Autosheet.](https://autosheet.com/docs/gpt-for-sheets/setup/open) 2. In the sidebar, select **GPT functions**. 3. Expand **Model settings** and open the model switcher. ![Open the model switcher](https://autosheet.com/docs/article-images/gpt-for-sheets/model-settings-switcher-functions.png) 4. Select a vision model. ![Select a vision model](https://autosheet.com/docs/article-images/gpt-for-sheets/model-switcher-vision.png) 5. In the spreadsheet, select a cell and write your formula: ``` =GPT_VISION("Extract the coffee brand from this image. Normalize the brand name to title case.",C9) ``` ![A GPT_VISION formula](https://autosheet.com/docs/article-images/gpt-for-sheets/result-vision-function-switcher.png) The formula uses the vision model selected in the model switcher. For instructions on how to write formulas with the GPT_VISION function and set the image resolution, see the [GPT_VISION](https://autosheet.com/docs/gpt-for-sheets/gpt-functions/reference#gpt_vision-function) reference. ### Specifying the model in the formula 1. [Open GPT for Sheets – Autosheet.](https://autosheet.com/docs/gpt-for-sheets/setup/open) 2. In the spreadsheet, select a cell and write your formula, specifying a vision model directly in the formula: ``` =GPT_VISION("Extract the coffee brand from this image. Normalize the brand name to title case.",C9, , , "gpt-4.1") ``` ![A GPT_VISION formula with the model parameter](https://autosheet.com/docs/article-images/gpt-for-sheets/result-vision-function.png) The formula uses the vision model regardless of what's currently selected in the model switcher. For instructions on how to write formulas with the GPT_VISION function and set the image resolution, see the [GPT_VISION](https://autosheet.com/docs/gpt-for-sheets/gpt-functions/reference#gpt_vision-function) reference. ## Host images on Google Drive for AI processing To process an image in GPT for Sheets – Autosheet, you need the direct URL of the image and the URL must be publicly accessible. To host an image on Google Drive so that Autosheet can access and process it: 1. Upload the image to Google Drive. 2. Share the image publicly using the **Anyone with the link** option. For more information, see [Google Drive Help](https://support.google.com/drive/answer/2494822). > **Tip** > > To prevent other users from commenting or editing the image, set the access role to **Viewer**. 3. Copy the link. 4. Get the image file ID from the link URL, and build the following new URL with the ID: ``` https://drive.google.com/uc?id= ``` For example: ``` # Original share URL from "Copy link" https://drive.google.com/file/d/1dTZrIkAHsV_hKnvL2KjH1pfcOOSpYVux/view # Direct image URL https://drive.google.com/uc?id=1dTZrIkAHsV_hKnvL2KjH1pfcOOSpYVux ``` You now have a direct image URL that you can access and process in Autosheet.