Google today has unveiled the new Gemma 3 open source ielligence model for developers. The Jama 3 model is actually the third version of Google’s Open Source. The company claims their new model is the world’s best model to run on only one GPU or CPU.
According to Google, the GEMMA 3 open -air model will be available to developers who iend to build artificial ielligence applications. It also supports 4 languages and even has the ability to analyze text, images and short videos.
The technology used in the construction of this Open Source model is exactly the same as the technology of making Jina’s artificial ielligence chats. Coext Window has grown in the Jama 3 model and has reached 128,000 token than 80,000 token in Jama 2; So it can understand more information and more sophisticated requests.
Google Gemma 3 Artificial Ielligence Model capabilities
Google claims that the GEMMA 3 is the best model for single-core use in the world and can be able to overcome competitors such as LLAMA models of Meta, Deepsic and Openai. Google also says users can launch this model only on a GPU or TPU. In addition, Jama 3 has the optimized capabilities for implemeing Nvidia graphics processors and proprietary artificial ielligence hardware.
The Gemma 3’s visual encoder is improved compared to previous models, and this model now supports high resolution images and other than square dimensions. The new ShieldGemma 2 image safety classification also helps users filter the input and output of images for viole or sensitive cases.
The popularity of Open Source Artificial Ielligence models among users and developers has already been in a state of uncertaiy, but the stunning growth and rapid popularity of Deepsic showed that the artificial ielligence market is thirsty for open -ended models with low hardware needs. Of course, because of safety issues, Google has restricted the use of GEMMA openings in specific areas.
The Jama 3 model is now available to developers through Google’s AI Studio. Developers can also download it through Kaggle or Hugging Face.
