Google has recently introduced its latest AI model, Gemini, featuring three tiers: Nano, Pro, and Ultra.
Among these, Gemini Ultra stands out as the most powerful, surpassing OpenAI’s GPT in a wide array of industry-standard benchmarks, according to internal sources within the company.
Insiders reveal that Gemini Ultra excels in 30 out of 32 widely recognized academic benchmarks for testing Large Language Models (LLM), outperforming even ChatGPT.
This development is significant for Google, as it seeks to address criticisms of trailing behind AI competitor OpenAI, whose GPT model has gained acclaim as the most dominant and widely used in the industry.
Insider reports highlight Gemini Ultra’s remarkable achievement in scoring 90% in the MMLU (massive multitask language understanding) benchmark.
SOPA Images / Contributor/ SOPA Images / Contributor
Notably, it becomes the first model to outperform human experts in this benchmark, which encompasses 57 subjects covering areas such as math, physics, history, law, medicine, and ethics, evaluating both world knowledge and problem-solving abilities.
Google has already integrated Gemini Nano and Pro AI models into its products, with Nano powering AI features in the Pixel 8 Pro smartphone, and Pro accessible through the Bard chatbot.
The release of Gemini Ultra is anticipated in early 2024. Moreover, Gemini Pro will be made available to enterprise customers and AI studio developers through Google’s Vertex AI program.
Eli Collins, the Vice President of Product at DeepMind, the Google division responsible for developing the AI platform, emphasizes Gemini Ultra’s capacity to comprehend “nuanced” information across diverse formats, including text, images, audio, and code.
Read More: iPhone 16: release date, latest leaks, price predictions and more
Collins also mentions that some of the data used in creating the application was derived from publicly available web content, though specific origins were not explicitly disclosed by the company.