Google has launched its Gemma 3 portfolio of lightweight LLMs, which include four model variants optimized to run efficiently across various devices, from workstations to smartphones. With performance surpassing competitors like OpenAI and DeepSeek, these models support over 140 languages and offer advanced text and visual reasoning capabilities. They also feature robust security measures and compliance with Google's data governance standards. Additionally, the release includes ShieldGemma 2, focusing on image verification and further enhancing the model's utility in ensuring safe AI applications.
Google's Gemma 3 LLMs introduce lightweight AI models that run on a single GPU, enabling broader access to advanced AI functionalities across various devices.
These new models come in four variants—ranging from 1 billion to 27 billion parameters—catering to developers' hardware capabilities for efficient AI application development.
The Gemma 3 LLMs reportedly outperform competitors like OpenAI's Llama-40B and DeepSeek-V3, with enhanced capabilities including multimodal reasoning and a token context window of 128,000.
In addition to the new lightweight LLMs, Google has launched ShieldGemma 2 for image verification, enhancing AI output safety and compliance with comprehensive data governance.
Collection
[
|
...
]