ETtech Explained | Google rolls out ChatGPT rival Gemini: its smartest AI model yet

Search major Google on Wednesday launched Gemini, said to be its ‘largest and most capable AI model’ yet. The model can even tell apart a real-life blue rubber duck from a drawing of a duck, Google demonstrated in a video. As the global AI race heats up, here’s all you need to know about Gemini and what it means.

What does Gemini bring to the table?

Gemini is said to be Google’s most flexible large language model (LLM) yet, capable of running on everything from data centres to mobile devices.


“Its state-of-the-art capabilities will significantly enhance the way developers and enterprise customers build and scale with AI,” Demis Hassabis, CEO and cofounder of Google DeepMind, said in a blog post, writing on behalf of the Gemini team.

Gemini has been optimised for three versions. Gemini Nano, optimal for mobile devices; Gemini Pro, which is built for scaling across a wide range of tasks; and Gemini Ultra, the largest model capable of undertaking highly complex tasks.

Where is it being rolled out?

Gemini is being rolled out through Google products.

Google’s generative AI chatbot Bard is getting its ‘biggest upgrade yet’ through integration with a fine-tuned version of Gemini Pro, to be available in English in more than 170 countries. Bard will be upgraded further with Gemini Ultra early next year.

Google smartphone Pixel 8 Pro will be able to run Gemini Nano, which will power new features such as ‘Summarise’ in the Recorder app. It will also roll out in Smart Reply in Gboard, beginning with WhatsApp, with more messaging apps to be enabled next year.

Gemini will be integrated with Google Search, Ads, Chrome and Duet AI in the coming months.

From December 13, developers and enterprise customers can access Gemini Pro via the Gemini API in Google AI Studio or Google Cloud Vertex AI, Google said.

Gemini Ultra is being further refined and tested for safety before developers and enterprises can access it early next year.

Also read | GenAI models not always factual, issue needs to be worked on, fixed: Google’s Jeff Dean


How is it different from other AI models?

Gemini is built to be natively multimodal, pre-trained from the start on different modalities, Google said. This means it can comprehend text, audio, image, video, and computer code simultaneously.

On the other hand, competitors including rival OpenAI’s ChatGPT are largely text-based and rely on plug-ins for image analysis and accessing the web. ChatGPT, for instance, also relies on Whisper and Dall-e to process images and audio.

It looks like Gemini has been oriented towards Google products, while models like ChatGPT and Meta’s Llama are more service-oriented, available to third-party developers for applications, tools and services, according to ZDNet.

Does this mean Google is ahead in the gen-AI space?

Gemini takes things a notch up in the race the global tech giants are locked in to come out with the most advanced AI.

According to Google, Gemini surpasses ‘state-of-the-art’ AI models on a range of benchmarks, including text and coding.

It is said to outperform GPT-4 on 30 out of 32 benchmark tests, including in reasoning and image understanding, while Gemini Pro outperforms GPT-3.5, which powers the free version of ChatGPT, in six out of eight tests.

Gemini has reasoning capabilities that enable it to ‘think more carefully’ before answering questions, Google said in a blog post.

It added that Gemini is the “first model to outperform human experts on MMLU (massive multitask language understanding), which uses a combination of 57 subjects such as math, physics, history, law, medicine and ethics for testing both world knowledge and problem-solving abilities”.