Google Introduces Gemini, Its Most Advanced AI to Date; Outperforms OpenAI’s GPT-4 in Benchmark Tests

Google has released Gemini, its next-generation major language model that powers the likes of the Google Bard chatbot and the company’s Pixel 8 Pro smartphone. The search giant calls Gemini its “most capable and most common model,” and thus replaces the PaLM 2. It’s built to be multimodal – meaning it can understand and use different types of data, such as text, code, images, and even videos.

“This new era of models represents one of the biggest science and technology efforts we’ve ever made as a company. I’m truly excited about what’s to come and the possibilities that Gemini will open up for people everywhere,” Sundar Pichai said in the release.

Gemini is available in three sizes for different platforms

In addition to setting new records in benchmarks, Gemini is also flexible. Google says it can work in data centers; At the same time, it is designed to work on something as pocket-friendly as a mobile phone.

Google Gemini 1.0 is available as Gemini Ultra, Gemini Pro (powered by Google Bard), and Gemini Nano (for phones like the Pixel 8 Pro). This is expected to help developers and enterprise customers easily scale and build with AI.

Beats OpenAI’s GPT-4 benchmarks

We also have a direct head-to-head comparison with GPT-4 powered by ChatGPT Plus and Microsoft Bing AI Copilot. According to Google, Gemini outperforms GPT-4 in text-based comparisons in general, reasoning, math, and code categories, and repeats the same in multimodal comparisons in image, video, and audio categories.

In addition, Gemini Ultra achieved a score of 59.4 percent on the new MMMU benchmark, which “consists of multimodal tasks that span different domains and require deliberate reasoning,” Google said. It also outperforms previous models in image comparisons without the help of Object Character Recognition Systems (OCR).

Gemini is already coming to market in Google products

You can experience Gemini’s capabilities with Google Bard to see how Gemini Pro performs compared to PaLM 2. Google is calling this the biggest update to Bard since its launch earlier this year.

Additionally, Google Pixel 8 Pro users would also be able to access Gemini Nano through apps like Recorder, Gboard and even WhatsApp, with support for chat apps coming later.

Google also says that Gemini will soon be available in services like search, ads, Chrome and Duet AI. And for developers and enterprises, the company makes Gemini Pro available through the Gemini API in Google AI Studio or Google Cloud Vertex AI.

Google Bard “Advanced” coming soon

Gemini Ultra will take effect on Google Bard Advanced from the beginning of 2024, and Google will also take a more responsible approach to the Ultra version by performing trust and security checks before its wide availability. It will also be available to select developers, customers and partners, as well as “security and liability experts” in 2024.