tiprankstipranks
Market News

Google (GOOGL) Launches Its Gemma 3 AI Model That Can Run on a Single Chip

Story Highlights

Tech giant Google has launched its latest open-source AI model, Gemma 3, which can run on a single GPU or TPU.

Google (GOOGL) Launches Its Gemma 3 AI Model That Can Run on a Single Chip

Tech giant Google (GOOGL) has launched its latest open-source AI model, Gemma 3, which can run on a single graphics processing unit (GPU) or tensor processing unit (TPU). These lightweight models are built using the same research and technology that powers Google’s Gemini 2.0 models. According to Alphabet’s (GOOG) (GOOGL) CEO Sundar Pichai, Gemma 3 models are incredibly efficient, with the largest 27B model running on just one H100 GPU.

The efficiency and flexibility of Gemma 3 models make them an attractive option for developers. They come in various sizes (1B, 4B, 12B, and 27B) to accommodate different hardware and performance needs. Additionally, Gemma 3 offers support for over 35 languages out-of-the-box and pre-trained support for over 140 languages, which makes it a versatile tool for AI applications.

In terms of performance, Gemma 3 has already shown some promising results. In preliminary evaluations, it outperformed other AI models, such as Meta Platforms’ (META) Llama-405B and OpenAI’s o3-mini. Google is also exploring new ways to put Gemma 3 to work. For instance, the company just rolled out ShieldGemma 2, a tool that uses Gemma 3 to help keep online images safe. Additionally, Google plans to support researchers who want to build on Gemma 3 through its Academic Program by offering Google Cloud credits to help their work.

Is Google Stock a Good Buy?

Turning to Wall Street, analysts have a Moderate Buy consensus rating on GOOGL stock based on 25 Buys and 11 Holds assigned in the past three months. After an 18% rally in the past 12 months, the average GOOGL price target of $215.88 per share implies 31.4% upside potential from current levels.

See more GOOGL analyst ratings

Questions or Comments about the article? Write to editor@tipranks.com