Google Open-Sources Powerful Edge Model: 2B Parameters Surpass GPT-3.5 Turbo, Runs Fast on iPhone 15 Pro
Google open-sources Gemma 2 models: 2B version surpasses GPT-3.5 Turbo, runs fast on iPhone 15 Pro. Discover ShieldGemma and Gemma Scope for enhanced AI safety and transparency.
At the end of June this year, Google open-sourced the Gemma 2 models in 9B and 27B versions. Since its debut, the 27B version quickly became one of the highest-ranked open models in the LMSYS Chatbot Arena, outperforming larger models in real conversation tasks.
Now, just over a month later, Google has furthered its commitment to responsible AI by enhancing the safety and accessibility of these models with new developments.
The latest updates include:
Gemma 2 2B: A lightweight version with built-in safety features, balancing performance and efficiency.
ShieldGemma: A security classifier model built on Gemma 2 to filter AI inputs and outputs, ensuring user safety.
Gemma Scope: A model interpretability tool offering unparalleled insights into the model's workings.
Notably, Gemma 2 2B stands out, scoring 1130 in the LMSYS Chatbot Arena with just 2 billion parameters, surpassing GPT-3.5-Turbo (0613) and Mixtral-8x7b. This makes it an ideal choice for edge devices.
Keep reading with a 7-day free trial
Subscribe to AI Disruption to keep reading this post and get 7 days of free access to the full post archives.