Google has introduced Gemini 1.0, an artificial intelligence model powered by LLM, which was unveiled at the recent Google I/O event. Google claims that the test results have shown Gemini outperforming GPT-4 in almost every aspect.
Gemini is a multimodal AI model that can process various types of data, including text, code, audio, images, and videos. In its 1.0 version, Google has released three sizes of Gemini. The largest is Gemini Ultra, designed for complex tasks. Gemini Pro comes next, offering versatility in its functionality. Lastly, Gemini Nano is a high-performance model optimized for running on mobile phones.
Google has showcased Gemini’s performance on different test suites, such as MMLU for various subject questions, GSM8K for mathematical evaluations, and HumanEval for Python code writing. In almost every test, Gemini has outperformed GPT-4, even in visual tests where it surpasses GPT-4V. Additionally, Gemini has shown superiority over the specialized Whisper-3 model when converting audio data into text.
Google plans to release Gemini Pro to Bard users initially, limited to the English language. However, they intend to expand language support and input options in the future. For Pixel 8 Pro users, Gemini Nano will be available next year, providing voice summarization and text assistance features. Other services will also incorporate Gemini.
TLDR: Google has introduced Gemini 1.0, an AI model that outperforms GPT-4 in various tests, supporting text, code, audio, images, and videos. Gemini comes in three sizes, with Gemini Pro releasing first to Bard users and Gemini Nano for Pixel 8 Pro users next year. Google plans to expand language support and input options in the future.
Leave a Comment