Google has introduced several time-saving and fun to use AI tools at its recent Google I/O event 2025 at Mountain View, California. Let us know about what each Google AI tool can do for you.
Note: Modal means the type of input like text, image, audio and video.
Gemini 2.5 Flash
Gemini 2.5 Flash AI model offers capabilities like answering prompts, creating summaries from files and photos. It is multimodal and allows you to converse with native audio in 24 languages seamlessly. Gemini 2.5 Flash is one step lower than Gemini 2.5 Pro. So, it is cost-effective and saves your budget. It allows long text prompts for better describing the problem or a project, or an expectation.
Gemini 2.5 Pro
It is the most advanced AI model from Google. A reasoning mode called DEEPTHINK constantly improves Gemini 2.5 Pro model. Gemini 2.5 Pro can do better in Advanced Coding, Maths and Science works. You can converse with native audio in 24 languages seamlessly. Its multimodal capability allows you to enter prompts in various modals.
Google AI Studio
Google AI Studio is like a portal. It can do 4 different types of tasks.
- Media Generation: It can generate image, audio and video samples using prompts. Using Imagen-4 model, it can generate near to realistic images.
- Stream: It can see through your Mobile or Laptop camera or things on a Screen and can answer your questions like a human being. It can judge your surroundings very well.
- Building AI Apps with Gemini
- Responding to Chats using multimodal inputs including YouTube videos. It can summarize or answer things.
Vertex AI Gemini API
This Vertex AI Gemini API is useful to develop Gemini applications. Gemini applications are similar to Gemini app that responds to various prompts. It is now available as Google Gen AI SDK. Google offers Vertex AI Model Builder to build and train custom models. There is also a library of enterprise models available as Vertex AI Model Garden for kickstart.
Gemini Developer API
This Gemini Developer API is useful to develop enterprise applications featuring generative AI. These applications run on Google Cloud Platform preferably. It is also available as Google Gen AI SDK now.
Firebase Studio
Firebase Studio is a full fledged IDE (Integrated Development Environment) that works in any browser just like that. You can develop frontend, backend and mobile applications keeping firebase as the database. You can migrate to any other databases or backend also later.
It supports many frameworks and programming languages like Go, Python Flask, Java, .NET, Angular, Flutter, Android Studio Cloud, NextJS, Node Express, Gemini API, Genkit, Astro, Mongo DB, React Native + Expo. You can import project from local system or from Github. The Firebase Studio supports Gemini Code Assist with contextual options.
Though this Firebase Studio is a browser-based tool. It supports debugging, testing and emulation just like a regular IDE. You can test your android or flutter or any other webapp without leaving the page. You can share the project URL to other collaborators for working like a team.
Firebase Studio effortlessly offers Code Completion and Code Generation with the help of Gemini AI. Just select some code and right-click to get Gemini contextual actions like below.
Gemini code assist
Gemini Code Assist is an AI assistant that helps in code completion, suggestion and generation. It is available for individuals as free while the paid versions are available as STANDARD and ENTERPRISE. This AI assistant is bundled along with IDEs such as IntelliJ, Android Studio and Visual Studio Code. The supported programming languages by the Gemini Code Assist are Bash, C, C++, C#, Dart, Go, GoogleSQL, Java, JavaScript, Kotlin, Lua, MATLAB, PHP, Python, R, Ruby, Rust, Scala, SQL, Swift, TypeScript and YAML.
Jules
Jules is a tool using Google Gemini AI for doing routine works like testing, version upgrading or downgrading and even fixing bugs. It is an asynchronous development agent. Jules takes input as your repository like a Github project. Jules works best with the programming languages like Java, Python, Go, Rust and JavaScript or TypeScript. You can test new features in a separate cloud based VM securely and decide whether to commit or not.
You can connect your Github's repository after logging into Jules.
Gemini Diffusion
Gemini Diffusion is an AI Model that works by text diffusion rather that auto regressive models. It is part of the Google DeepMind model. It works faster by accepting more tokens in a single go. It can generate solutions faster step by step and correcting each step before proceeding to the next step at generation stage itself. It achieved good benchmarks in bulk code generation and solving math problems.
Veo
Google Veo is an AI Video Generation tool that can be accessed from different product like Flow, Google Vids, Google AI Studio, Vertex AI Studio and Gemini. Google Flow leads the show with the features like creation of cinematic clips, scenes and stories with just text prompts and reference images and videos. Flow is an AI film making tool that can be used by professionals to create creative movie quality videos with features like consistent characters, camera controls, native audio and motion controls.
Gemini Robotics
Gemini Robotics is an AI Model which is under continuous development to bring Robots into the day to day activities of human beings that we mostly see only in movies. Gemini Robots can understand the real world with good reasoning skills.
Google showcases some tasks being done by Gemini Robots like Packing a lunch bag, closing the zipper of lunch bag, playing with English alphabet toys, doing origami and arranging the letters and more. It enables fine motoring skills for precise control and pressure.
Project Astra
Project Astra is a video understanding AI Prototype. It powers some of the latest features of Gemini Live. It can understand your Android Phone apps, live surroundings when given access to the camera and the AI Glasses platform. It communicates with many of the apps to do works on your behalf or wait for the final confirmation from you. It can listen, understand and reply in 24 languages without any delay with the help of native audio support.
Project Astra is in collaboration with some services like Aira to help blind people navigate through real world.
Gemini Live
Gemini Live is an AI Model that can see, listen and respond to text, image, audio, video and live footage in an exceptional way. You can get directions (not that google maps thing) and instructions from what you see around you like a human to complete your project in a more creative way. Gemini Live excels in contextual understanding even if you change topics in the middle of a conversation with it. It can adapt your way of seeing and understanding things around you.
AlphaProof
AlphaProof and AlphaGeometry are AI models that can solve mathematical problems. AlphaProof can solve reasoning questions well, while AlphaGeometry-2 excels in solving geometrical problems with its neuro-symbolic engine. AlphaProof was able to reach the Gold Medal stage in IMO (International Mathematical Olympiad) along with human participants. This revolutionises the learning process for the students and schools.
AI Co-Scientist
This Google AI Co-Scientist can assist scientists in creating great hypotheses and proposals in considerably less time, accelerating the progress.
AI Co Scientist model contains components or agents like Generation Agent, Reflection Agent, Evolution Agent, Proximity Check Agent, Meta Review Agent, Research Plan Configuration, Research Goal and Research Proposals Overview.
Alpha Evolve
Alpha Evolve is an AI model that is based on large language models (LLM). It can understand and summarize documents, generate code etc. It evolves and improves itself by solving algorithms and can generate computer program code. It can solve mathematical algorithms and create new algorithms.
AMIE (Articulate Medical Intelligence Explorer)
AMIE is an AI model that can have conversations with the patients like a clinician. It can generate a questionnaire based on the patient's input and past medical history. AMIE's diagnostic AI can create empathy and assurance in patients like real human beings. AMIE was able to achieve accuracy 26 out of 28 in the Clinician's perspective and 24 out of 26 in the patient's perspective.
AlphaFold 3
AlphaFold 3 can simulate the interactions of various proteins with other elements in biology. It can predict the behaviours of hundreds of proteins simultaneously, leading to a better understanding of diseases and improvements in the human body. This Alpha Fold 3 AI model aims at reducing diseases in crops, humans, and insects and increasing food safety for future generations.