What Are BYOM and Local AI in Foxit AI Assistant?

Quick Start

Last updated September 3, 2026 · 4 min read · Tier L1

By default, Foxit AI Assistant runs on Foxit-hosted models and draws from your subscription's shared credit pool. If you need your AI usage to stay on your own infrastructure, or you already have a paid model subscription you'd rather use, AI Assistant also supports Local AI (on-device) and Bring Your Own Model (BYOM) -- connecting any OpenAI-compatible model endpoint -- both configured from AI Settings.

Local AI and BYOM are both configured in one place: AI Settings, inside Foxit PDF Editor or Foxit PDF Reader (the desktop app, not Foxit Workspace). Here's exactly how to get there:

1. Open a document in Foxit PDF Editor or Foxit PDF Reader.
2. Click the AI Assistant tab in the ribbon at the top of the window.
3. Click the AI Settings icon (the gear, at the right-hand end of that ribbon).

This opens the AI Settings window, with a sidebar for General, Usage, Connectors, Skills, and Models. Local AI and BYOM are both configured from the Models page.

Local AI runs a small language model (SLM) directly on your device -- currently Phi-4 Mini or Qwen3-4B, downloaded and run via Microsoft Foundry. Your documents and prompts never leave your device, which makes it a fit for sensitive content that can't go to a cloud service. It also works offline and has no cloud API cost.

BYOM connects Foxit AI Assistant to a model you already have access to, hosted by you or a third party, instead of using Foxit's built-in models. The connection must be OpenAI-compatible.

• Data privacy -- with Local AI, no document content or prompt is sent to a third-party or Foxit-hosted service. With BYOM, your data goes only to the model endpoint you connect -- not to Foxit.
• No double-paying for AI -- if your organization already pays for a model subscription, BYOM lets you use it instead of a separate AI subscription.
• Credits aren't consumed -- usage through Local AI or a connected BYOM model does not draw from your Foxit AI credit pool. Credits are only metered when you use Foxit's own hosted (Cloud) models.

1. Open AI Settings and go to Models (see Where do I find AI Settings? above).
2. Select the On-Device tab.
3. Download one of the available small language models under Models (currently Phi-4 Mini or Qwen3-4B), and choose whether it runs on your GPU, CPU, or NPU.
4. Toggle the model on. AI Assistant now processes requests on your device instead of in the cloud.

1. Open AI Settings and go to Models (see Where do I find AI Settings? above).
2. Select the Custom tab, then click Add Models.

1. Fill in your model's connection details: Base URL, API Key, Model (the model identifier), and a Display Name to show in Foxit's model picker. Select whether the model supports Tools and/or Vision.

1. Save. Your model now appears in the Models list and can be turned on for use in AI Assistant.

Your connection must point to an API that follows the OpenAI API format -- meaning it accepts requests and sends back responses shaped the same way OpenAI's API does. This isn't the same as connecting to OpenAI itself: it's become a widely-used industry standard, so many self-hosted model servers and other third-party AI providers also use this same format, and any of them can work as long as they follow it. Foxit doesn't manage or vet the third-party or self-hosted endpoints you connect this way -- the security, reliability, and data handling of that endpoint are the responsibility of whoever hosts it, not Foxit.

• How Do I Check My AI Credit Usage in Foxit Workspace?
• AI Workspace Licensing, Access, and Availability