Opera, the Norwegian software company known for the web browser of the same name, announced a few days ago that it will add to its browser Opera One experimental support for 150 variants of local large language models (LLM) from about 50 model families.
Local LLMs allow users to access AI models directly from their computer without the processing ever leaving their device. This means that latency is lower because the data used is not sent over the Internet. Since they do not require an internet connection, this also means that they cannot be used to train other language models.
The new feature is part of the AI Feature Dropsde Opera program and is accessible as part of the Opera browser developer channel, as explained by the company in a press release. Interested users will need to update to the latest version of Opera Developer to access local LLMs.
“The introduction of local LLMs in this way allows Opera to start exploring ways to create experiences and knowledge in the emerging field of local AI,” said Krystian Kolondra, Executive Vice President of Browsers and Gaming at Opera.
Among the local models compatible with Opera One are Llama by Meta, Gemma by Google, Mixtral by Mistral AI, and Vicuna. Most of the local LLMs require between 2 and 10 gigabytes of local storage space per variant and replace Aria, the native AI of the Opera browser.
To download and activate one of the new variants, users will have to open the Aria chat, choose “local mode” from a dropdown menu, and select a model to download from the settings. From there, it will be possible to download and then activate several models.
