Opera has opened an early preview of WebAI, a technology that lets AI models run directly in its browser. Processing takes place on the user's computer, removing the need to send data to a cloud AI service for those tasks.

The initial rollout is in the flagship Opera One browser. Opera plans to bring the capability to Opera GX, aimed at gamers and streamers, and Opera Neon, where AI agents carry out user tasks. Users must actively enable the feature.

Which models WebAI supports

WebAI is a proposed web standard that gives websites and extensions access to local AI through the browser. Opera's implementation is not tied to a single provider: the company cites Google's Gemma models and, on compatible Macs, Apple's on-device foundation model among the available options.

The release builds on Opera's earlier work with local models. In 2024, the company introduced built-in access in an experimental developer version of its browser.

Open-weight models are central to this approach. They are generally free to download and use, while access to their parameters gives developers more scope to adapt their behavior. However, open weights do not automatically make a model fully open-source software: licensing terms may still impose restrictions.

What changes for users and developers

WebAI supports text generation, content classification and identifying meaningful relationships between pieces of information. Developers can integrate these capabilities into websites and extensions without operating separate cloud infrastructure or paying an external AI service for every request.

For users, this offers access to local AI features without a cloud model subscription. Opera says information processed by local AI remains on the device. Users can also decide which websites are allowed to access the models.

Privacy also depends on where data is processed

As AI assistants gain access to more personal information, understanding where processing happens becomes more consequential. Local models reduce the need to send that information to an outside provider and give users more control. Running a model on a personal computer does not, however, remove every privacy risk.

There is also a hardware trade-off: the device must handle the computational workload. Opera cautions that response speed will depend on the computer's hardware, and local AI can affect its overall performance.

Opera has opened an early preview of WebAI, a technology that lets AI models run directly in its browser. Processing takes place on the user's computer, removing the need to send data to a cloud AI service for those tasks.

The initial rollout is in the flagship Opera One browser. Opera plans to bring the capability to Opera GX, aimed at gamers and streamers, and Opera Neon, where AI agents carry out user tasks. Users must actively enable the feature.

Which models WebAI supports

WebAI is a proposed web standard that gives websites and extensions access to local AI through the browser. Opera's implementation is not tied to a single provider: the company cites Google's Gemma models and, on compatible Macs, Apple's on-device foundation model among the available options.

The release builds on Opera's earlier work with local models. In 2024, the company introduced built-in access in an experimental developer version of its browser.

Open-weight models are central to this approach. They are generally free to download and use, while access to their parameters gives developers more scope to adapt their behavior. However, open weights do not automatically make a model fully open-source software: licensing terms may still impose restrictions.

What changes for users and developers

WebAI supports text generation, content classification and identifying meaningful relationships between pieces of information. Developers can integrate these capabilities into websites and extensions without operating separate cloud infrastructure or paying an external AI service for every request.

For users, this offers access to local AI features without a cloud model subscription. Opera says information processed by local AI remains on the device. Users can also decide which websites are allowed to access the models.

Privacy also depends on where data is processed

As AI assistants gain access to more personal information, understanding where processing happens becomes more consequential. Local models reduce the need to send that information to an outside provider and give users more control. Running a model on a personal computer does not, however, remove every privacy risk.

There is also a hardware trade-off: the device must handle the computational workload. Opera cautions that response speed will depend on the computer's hardware, and local AI can affect its overall performance.

Opera has opened an early preview of WebAI, a technology that lets AI models run directly in its browser. Processing takes place on the user's computer, removing the need to send data to a cloud AI service for those tasks.

The initial rollout is in the flagship Opera One browser. Opera plans to bring the capability to Opera GX, aimed at gamers and streamers, and Opera Neon, where AI agents carry out user tasks. Users must actively enable the feature.

Which models WebAI supports

WebAI is a proposed web standard that gives websites and extensions access to local AI through the browser. Opera's implementation is not tied to a single provider: the company cites Google's Gemma models and, on compatible Macs, Apple's on-device foundation model among the available options.

The release builds on Opera's earlier work with local models. In 2024, the company introduced built-in access in an experimental developer version of its browser.

Open-weight models are central to this approach. They are generally free to download and use, while access to their parameters gives developers more scope to adapt their behavior. However, open weights do not automatically make a model fully open-source software: licensing terms may still impose restrictions.

What changes for users and developers

WebAI supports text generation, content classification and identifying meaningful relationships between pieces of information. Developers can integrate these capabilities into websites and extensions without operating separate cloud infrastructure or paying an external AI service for every request.

For users, this offers access to local AI features without a cloud model subscription. Opera says information processed by local AI remains on the device. Users can also decide which websites are allowed to access the models.

Privacy also depends on where data is processed

As AI assistants gain access to more personal information, understanding where processing happens becomes more consequential. Local models reduce the need to send that information to an outside provider and give users more control. Running a model on a personal computer does not, however, remove every privacy risk.

There is also a hardware trade-off: the device must handle the computational workload. Opera cautions that response speed will depend on the computer's hardware, and local AI can affect its overall performance.

Opera has opened an early preview of WebAI, a technology that lets AI models run directly in its browser. Processing takes place on the user's computer, removing the need to send data to a cloud AI service for those tasks.

The initial rollout is in the flagship Opera One browser. Opera plans to bring the capability to Opera GX, aimed at gamers and streamers, and Opera Neon, where AI agents carry out user tasks. Users must actively enable the feature.

Which models WebAI supports

WebAI is a proposed web standard that gives websites and extensions access to local AI through the browser. Opera's implementation is not tied to a single provider: the company cites Google's Gemma models and, on compatible Macs, Apple's on-device foundation model among the available options.

The release builds on Opera's earlier work with local models. In 2024, the company introduced built-in access in an experimental developer version of its browser.

Open-weight models are central to this approach. They are generally free to download and use, while access to their parameters gives developers more scope to adapt their behavior. However, open weights do not automatically make a model fully open-source software: licensing terms may still impose restrictions.

What changes for users and developers

WebAI supports text generation, content classification and identifying meaningful relationships between pieces of information. Developers can integrate these capabilities into websites and extensions without operating separate cloud infrastructure or paying an external AI service for every request.

For users, this offers access to local AI features without a cloud model subscription. Opera says information processed by local AI remains on the device. Users can also decide which websites are allowed to access the models.

Privacy also depends on where data is processed

As AI assistants gain access to more personal information, understanding where processing happens becomes more consequential. Local models reduce the need to send that information to an outside provider and give users more control. Running a model on a personal computer does not, however, remove every privacy risk.

There is also a hardware trade-off: the device must handle the computational workload. Opera cautions that response speed will depend on the computer's hardware, and local AI can affect its overall performance.