Skip to main content
Brave Leo is the AI assistant built into Brave Browser. Leo’s Bring Your Own Model (BYOM) feature can connect directly to Venice’s OpenAI-compatible chat completions endpoint, letting you use Venice models for chat and page-aware assistance without installing an extension.

Browser Native

Chat, summarize pages, and explain selected text in Brave

Direct Connection

Send BYOM requests directly from Brave to Venice

Privacy Choice

Choose a Venice model with the privacy tier you need

​
Prerequisites

Brave’s BYOM feature is available on desktop. The model ID must match a Venice model ID exactly.

​
Setup

1

Open Leo settings

Open brave://settings/leo-ai, or open Leo from Brave’s sidebar, select the settings icon, and navigate to Leo.
2

Add a custom model

Scroll to Bring your own model and select Add new model.
3

Configure the Venice endpoint

Enter the following values:If Brave displays Context size, set it to the context window supported by your selected model. You can also add an optional system prompt.Enter the API key itself without a Bearer prefix. Brave adds the authorization header to each request.
4

Save and select Venice

Select Add model, then set Venice AI as the default model for new conversations or choose it from Leo’s model selector.
5

Test the connection

Open Leo on a webpage and send a prompt such as:
Leo sends the conversation and relevant page content directly to the Venice endpoint.
Brave does not expose Venice-specific request-body parameters in BYOM. To enable Venice web search, append a model feature suffix to the model ID in Model request name:
To always enable search and include citations, use:
Web search has additional usage-based pricing. See API pricing for current rates.

​
Privacy

When you use a remote BYOM endpoint, Brave connects directly to that provider. Requests bypass Brave’s servers, so Brave’s proxy and server-side retention protections do not apply to those requests. Venice does not store or log prompt and response content at its proxy. The selected model determines the runtime privacy tier:
  • Private models use zero-data-retention infrastructure.
  • Anonymized models hide identifying information from the upstream provider, but the provider can process the prompt.
  • TEE and E2EE models add hardware-backed isolation or client-side encryption where supported.
Check the model’s privacy label in the text model catalog and read Venice privacy before sending sensitive information. Brave may keep Leo chat history in local browser storage. You can clear Leo data, use a temporary chat, or disable local chat history in Brave’s Leo settings.

​
Configuration notes

  • Brave requires the full endpoint path: https://api.venice.ai/api/v1/chat/completions.
  • Model IDs and availability change over time. Copy the current ID from the text model catalog.
  • Context size should not exceed the selected model’s supported context window.
  • Page summaries can send page content to Venice as part of the prompt.
  • Remote Venice requests require an internet connection and consume Venice API credits or your available DIEM allowance.

​
Troubleshooting

Confirm that the key comes from venice.ai/settings/api. Paste the raw key without Bearer and remove any leading or trailing spaces.
Copy the model ID exactly from the text model catalog. If you added a feature suffix, first test the model without the suffix.
Confirm that the server endpoint includes /api/v1/chat/completions. Brave Leo expects an OpenAI-compatible streaming response.
Check Leo’s permission to use page content and make sure the current page is accessible to the browser. Internal Brave pages and some protected pages cannot be summarized.

​
Resources

Brave BYOM Guide

Official Brave setup and privacy details

Venice Text Models

Browse current model IDs and privacy tiers

Model Feature Suffixes

Enable web search and other Venice features

Generate an API Key

Create and securely store a Venice API key