Opera introduces local download and use of large language models

Opera has unveiled a feature, allowing users to download and utilize large language models (LLMs) directly on their computers.

Initially debuting for Opera One users receiving developer stream updates, this functionality grants access to a diverse set of over 150 models from more than 50 families, including prominent names like Llama from Meta, Gemma from Google, and Vicuna.

Through Opera's AI Feature Drops Program, users gain early access to this innovative capability, which leverages the Ollama open-source framework within the browser to execute these models locally.

Each LLM variant occupies more than 2 GB of local memory. Notably, Opera does not use space-saving algorithms when loading models. Jan Stendal, Opera's vice president, emphasized the importance of this development, noting that over time the size of models may decrease as they become more specialized for specific tasks.