Google unveils open source Gemma 3 model with 128k context window
1 min read
Summary
Google has launched Gemma 3, the latest version of its small model, offering robust performance for phones and laptops, while also reducing computing costs.
It has four different sizes for different tasks, and can analyse images, text and short videos, as well as supporting function calling in 140 languages.
To enhance performance, Google has developed quantised versions of Gemma 3, which it says delivers better results than larger language models from DeepSeek, OpenAI and Meta.
Google highlighted the benefits of smaller language models (SLMs) over larger language models (LLMs) for many uses, and said that in some cases, SLMs can deliver the same performance as LLMs but at a fraction of the cost.
The advantages include less overfitting, better contextual limitations, faster inference, lighter computational requirements and enhanced privacy.