# Microsoft Phi: Small Language Models With No Web Access

> Phi is Microsoft's family of small language models under the MIT license. Current Phi-4 models, and why Phi has no crawler or user agent to block.

Phi is the Microsoft family of small language models (SLMs). An SLM is a language model with a small number of parameters. Phi models are free to download under the MIT license. They do not browse the web and have no crawler or user agent. A site owner has nothing to allow or block for Phi.

## Phi models

The [Microsoft Phi page](https://azure.microsoft.com/en-us/products/phi) lists Phi-4, Phi-4-mini and Phi-4-multimodal as the current models. Phi-4-multimodal accepts text, audio and image input. The page also lists the older Phi-3.5, Phi-3, Phi-2, Phi-1.5 and Phi-1 series. Developers can get Phi models from Azure AI Foundry Models, Hugging Face and Ollama.

Microsoft published later Phi-4 variants on [Hugging Face](https://huggingface.co/microsoft). All the models in this table use the MIT license:

| Model | Published on Hugging Face |
|---|---|
| [phi-4](https://huggingface.co/microsoft/phi-4) | 2024-12-11 |
| [Phi-4-mini-instruct](https://huggingface.co/microsoft/Phi-4-mini-instruct) | 2025-02-19 |
| [Phi-4-multimodal-instruct](https://huggingface.co/microsoft/Phi-4-multimodal-instruct) | 2025-02-24 |
| [Phi-4-reasoning](https://huggingface.co/microsoft/Phi-4-reasoning) | 2025-04-09 |
| [Phi-4-reasoning-vision-15B](https://huggingface.co/microsoft/Phi-4-reasoning-vision-15B) | 2026-01-23 |

## Web access

A Phi model generates text from the input that it gets. It does not search the web or send requests to websites. If an app that uses Phi also reads web pages, the request comes from that app. Look for the user agent of that app in your logs.

Microsoft's assistant, Microsoft Copilot, gets web results from Bing. See the [Microsoft Copilot guide](/ai-search-engine/microsoft-copilot/) for how it uses your pages.

## Training data and your site

We found no Microsoft documentation of a separate crawler that collects web data for Phi. Bing has its own rule for AI training. Bing says that it does not use pages with the `noarchive` tag to train Microsoft generative AI foundation models ([Bing blog, September 2023](https://blogs.bing.com/webmaster/september-2023/Announcing-new-options-for-webmasters-to-control-usage-of-their-content-in-Bing-Chat)). Bing does not name the models that this rule covers. For the details, see the [Bingbot guide](/ai-crawler-bot/bingbot/).

---

[View the canonical page](https://aichatwatch.com/ai-chat-bot/phi/) · [Browse llms.txt](https://aichatwatch.com/llms.txt)