Google brings second-gen AI models to the Gemini mobile app

Earlier today, Google made a few notable AI announcements, at a time when the tech industry is peeling the layers of China’s DeepSeek AI and the search giant is staring at anti-trust heat in China. The latest from Google is an experimental version of Gemini 2.0 Pro, which is claimed to be the company’s latest and greatest so far.

“It has the strongest coding performance and ability to handle complex prompts, with better understanding and reasoning of world knowledge, than any model we’ve released so far,” says the company. This one raises the prompt context window to 2 million tokens, allowing it to ingest and comprehend massive inputs with ease.

On the more affordable side of things, Google is pushing the new 2.0 Flash-Lite model as a public preview. Focused on reduced costs and snappier performance, this one is now available in Google AI Studio and Vertex AI systems.

Please enable Javascript to view this content

What’s new Gemini mobile app?

For smartphone users, the Gemini app is now getting access to these AI upgrades. Starting today, the mobile application will let users pick between the new Gemini 2.0 Flash Thinking Experimental and Gemini 2.0 Pro Experimental models.

Currently ranked as No. 1 on the Chatbot Arena LLM Leaderboard — and ahead of OpenAI’s ChatGPT-4o and DeepSeek R1 — the Gemini 2.0 Flash Thinking Experimental model is a massive leap forward for a couple of reasons.

First, it can work with data pulled from apps such as YouTube, Google Maps, and Search. Based on your queries, this Gemini model can cross-check information from within those platforms and offer relevant answers.

Second, it comes with thinking and reasoning capabilities. To put it simply, you can see, in real-time, how this model breaks down your commands and puts together the information as a cohesive response.

The result, as Google puts it, is improved explainability, speed, and performance. It allows text and image input with support for up to a million token window, while the knowledge cut-off boundary is set to June 2024.

Next, we have the Gemini 2.0 Pro Experimental model, which is now available to folks who pay for a Gemini Advanced subscription. Google says this one is “exceptional at complex tasks,” particularly at chores such as maths problem-solving and coding.

This multi-modal AI model can also pull relevant data from Google Search, and combine it with enhanced world understanding chops to handle more demanding tasks. You can access these new Gemini 2.0 series models on the mobile app as well as the web dashboard.