GoogleOfficially launched its Gemma2The new member of the series - Gemma22B ModelThis model, with 2 billion parameters, has demonstrated excellent performance in a variety of hardware environments. In addition to the powerful model itself, Google also launched the ShieldGemma security classifier to filter harmful content and provided the Gemma Scope tool for researchers to analyze the model's decision-making process.
Gemma22B performed particularly well in the "Chatbot Arena" leaderboard, with a high score of 1130, successfully surpassing many competitors including GPT-3.5-Turbo-0613 and Mixtral-8x7b, and even competing with some models that are ten times larger than it. Thanks to the optimization of the NVIDIA TensorRT-LLM library, Gemma22B is highly efficient to deploy and is suitable for a variety of scenarios from edge devices to powerful cloud environments. This allows it to be used in data centers, local workstations, and edge AI applications.
Against the backdrop of the rapid development of AI technology, this new release by Google is undoubtedly an important breakthrough in the fierce competition. In this regard, Rebecca Weiss, executive director of ML Commons, said that as AI technology matures, the entire industry needs to invest more resources to develop high-performance security assessment tools, and looks forward to Google's continued efforts in this area.
In addition, the release of Gemma22B coincides with the launch of Meta's Llama3.1, which also surpasses OpenAI's GPT-4o in most benchmarks. OpenAI also released the more cost-effective GPT-4o mini last week, which is 30 times cheaper than GPT-40 and 60% cheaper than GPT-3.5Turbo.
CurrentLarge Language Models(LLM) Market competition is becoming increasingly fierce. If Google wants to continue to gain a foothold, it must increase its innovation efforts and launch a more competitive Gemma3.