Gemma 3: Google’s Fastest On-Device AI Model Yet

Cryptos: 21,331
Exchanges: 1,862
Market Cap: $2.66T 0.62%
24h Vol: $116,243,551,941 40.16%
Dominance: BTC: 61.3% ETH: 8.5%
Gas: 1 GWEI

Connect

Crypto News AI

Gemma 3: Google’s Fastest On-Device AI Model Yet

Written by Aaron S.,

Editor-In-Chief

Last Updated: March 12, 2025

Key Takeaways

Gemma 3 runs AI applications on devices without cloud reliance, optimizing efficiency;
Google claims Gemma 3 outperforms many larger AI models on a single GPU;
Despite being open-source, Gemma 3 has licensing limits, but Google offers cloud credits.

Gemma 3: Google’s Fastest On-Device AI Model Yet

BITDEGREE IN YOUR SOCIAL FEED

Google has introduced Gemma 3, the latest version of its lightweight artificial intelligence (AI) models designed to run directly on devices like laptops and phones.

Built on the same foundation as its larger Gemini AI, the new model is optimized for efficiency, allowing developers to create AI applications without relying on cloud processing.

According to Google, Gemma 3 delivers top performance among AI models that operate on a single GPU, outperforming competitors like Meta’s Llama-405B, OpenAI’s o3-mini, and DeepSeek-V3. Only DeepSeek R1 ranks higher.

What is Tezos? XTZ Cryptocurrency Easily Explained (ANIMATED)

Did you know?

Want to get smarter & wealthier with crypto?

Subscribe - We publish new crypto explainer videos every week!

The AI model can analyze text, images, and short videos, offering a 128k-token context window that enables it to process large inputs. It also includes built-in function calling and structured output.

Gemma 3 also introduces official quantized versions, which reduce precision levels to make the model smaller and improve processing speed.

While open-source developers have created similar modifications for other AI models, Google’s official versions are designed for maximum efficiency.

Google describes the Gemma 3 family as its “most advanced, portable, and responsibly developed open models yet”.

Despite its open-source branding, the model’s licensing restrictions remain unchanged, which limits how it can be used. However, Google is promoting its adoption by offering cloud credits, including $10,000 in research funding through the Gemma 3 Academic program.

Meanwhile, Alibaba Cloud recently introduced its new AI model, QwQ-32B. How does it compare to other AI models? Read the full story.