Google announces its latest model Gemma 3, which is based on Gemini 2.0 technology. Gemma 3 is said to outperform models from OpenAI, Meta, and DeepSeek.
Google introduces Gemma 3: a series of open models based on the same technology as Gemini 2.0. The company claims that Gemma 3 would surpass the AI models of other tech giants including Meta, OpenAI, and DeepSeek. The Gemma 3 models are primarily intended for developers who want to build AI applications. Gemma 3 is available in various sizes: 1B, 4B, 12B, and 27B.
AI Applications
“Gemma 3 delivers state-of-the-art performance for its size and outperforms Llama-405B, DeepSeek-V3, and o3-mini in preliminary evaluations of human preferences on the LMArena leaderboard,” according to the tech giant. Furthermore, the model supports more than 35 languages and offers pre-trained support for over 140 languages.
read also
Google launches Gemma 3: based on Gemini 2.0
Additionally, the model includes a context window of 128k tokens. This enables them to understand and process large amounts of information, according to Google. Moreover, the model can analyze text, images, and short videos.
Google states that Gemma 3 is intended for developers building AI applications that can run on various devices. The models are optimized for Nvidia and AMD GPUs, and for Google Cloud TPUs.