Select a model in Word

GPT for Word supports a range of models that you can use with your AI work in Word. To help you find the model that offers the best balance of accuracy, speed, and cost for your use case, see our recommended models.

The available models depend on:

• The API keys set for your current space. GPT for Word gives you access to a default set of models available without API keys. To access additional models, set an API key for one or more AI providers.

• The API endpoints set up for your current space. To access even more models, set up a connection to an Azure OpenAI endpoint or any OpenAI-compatible API endpoint.

• The model settings for your current space. Space owners and admins can control which models are available in GPT for Word.

Select a model without an API key

1. Open a Microsoft Word document.

2. On the ribbon, select Home > GPT for Excel Word.

   

   info

   If GPT for Excel Word is not visible on the ribbon, or if it's grayed out, select Home > Add-ins > My Add-ins > GPT for Excel Word. Learn more.

3. Expand the model switcher.

   

4. Under No API key needed, select a model.

   

GPT for Word now uses the selected model to generate responses.

Select a model with an API key

1. Open a Microsoft Word document.

2. On the ribbon, select Home > GPT for Excel Word.

   

   info

   If GPT for Excel Word is not visible on the ribbon, or if it's grayed out, select Home > Add-ins > My Add-ins > GPT for Excel Word. Learn more.

3. Expand the model switcher.

   

4. Under Use your API key or endpoint, select a model with a key icon.

   

GPT for Word now uses the selected model to generate responses.

Recommended models

As a general rule, OpenAI models are more reliable and less likely to decline prompts than models from other providers.

The following table lists our recommended models available without an API key. For more information about which add-on supports which models, see AI providers & models.

Model	Provider	Description
claude-4-sonnet	Anthropic	Great at content creation, summarization, and nuanced reasoning across a wide range of tasks. Recommended for most use cases.
gpt-4.1	OpenAI	Good overall balance of easy prompting, speed, quality, reliability, and cost.
gpt-4.1-mini	OpenAI	Fast, cost-effective model with strong reasoning and language capabilities. Follows instructions closely.
gpt-4.1-nano	OpenAI	Great for extremely high-volume but easy tasks.
gemini-2.5-flash	Google	Cost-effective model ideal for lightweight tasks and rapid responses. Supports reasoning, web search, and web scraping (enabled from the model settings).

To learn more about a model, hover over the model info icon in the model switcher. A tooltip opens with detailed information about the model.

What's next

Learn more about models:

• AI providers & models supported by GPT for Work

• LiveBench (a good model benchmark)