New free Google anti-gravity upgrade, Gemini 3.1 Pro. You are sleeping on one of the most powerful free AI tools available right now. And I get it, there are 100 tools dropping every week. Most of them are hype. Most of them don't actually change how you work. But what Google just did, this is different. They just upgraded their free agentic coding platform, Anti-gravity, with their newest and most capable model yet, Gemini 3.1 Pro. And by the end of this video, you'll know exactly what that means, why it matters, and how to get access right now for free. My name is Julian, and I help people learn and actually implement AI tools that save time and improve the way they work. So, let's get into it. First, let me set the stage. Google dropped anti-gravity back in November 2025. And if you missed that, don't worry. We're we're going to cover what it is. But what just happened on February 19th, 2026 is a big deal. Google announced Gemini 3.1 Pro and they shipped it directly into anti-gravity for developers to use right now. So before I tell you what Gemini 3.1 Pro

actually does, let me explain what anti-gravity is because if you haven't tried it yet, this is the reason you should. Anti-gravity is Google's free agentic development platform. And I want to be really clear about what Agentic means here because it's not just another buzzword. Traditional coding assistants, your co-pilots, your cursors, they autocomplete code. They sit next to you and help you write faster. You're still the one driving. Anti-gravity flips that entirely. You give it a task and autonomous agents plan it out, write the code, run it in the terminal, open the browser to test it, and come back to you with the result. You go from writing code line by line to managing a team of agents doing the work. The platform has two views. The editor view is your familiar IDE experience, think VS Code, where you can still get hands-on when you need to. The manager surface is where the shift happens. It's a mission control dashboard where you can spawn multiple agents, run them in parallel across different tasks, and just review what they produce. It's free for

individuals, available on Mac, Windows, and Linux. Now, I want to make sure you understand what just got plugged into it because that's the real story today. Let me talk about Gemini 3.1 Pro. This is Google's newest model just dropped. And the benchmark improvements are legitimately significant. On ARC AGI2, a benchmark that tests a model's ability to solve completely new logic patterns it's never seen before. Gemini 3.1 Pro scored 77.1%. Gemini 3 Pro scored 31.1% on the same test. That's more than double the reasoning performance of the previous version. Better reasoning means the agent doesn't just follow instructions, it figures things out. It handles ambiguity. It solves problems you didn't fully specify. That's a massive upgrade for real world use. But the Agentic benchmarks are what really caught my attention. On Terminal Bench 2.0, which tests how well a model handles aentic coding tasks in the terminal, 3.1 Pro scored 68.5% compared to 56.9% for Gemini 3 Pro. On MCP Atlas, a multi-step

workflow benchmark using the model context protocol. 3.1 Pro hit 69.2 2%. Gemini 3 Pro was at 54.1%. That's nearly a 15point jump. And on Browse Comp, which tests how well a model can search the web agentically to find real answers, 3.1 Pro scored 85.9%. Gemini 3 Pro was at 59.2%, these aren't small incremental upgrades, and every one of those benchmarks maps directly to something anti-gravity's agents actually do. Now, here's something specific I want you to know about because it changes how you use the model. Gemini 3.1 Pro introduces a new three tier thinking level system. In Gemini 3 Pro, you had two modes, low thinking and high thinking. Simple. With 3.1 Pro, Google added a medium setting, and they completely overhauled what high means. When you set the model to high thinking, it now behaves like a lighter version of Gemini deep think which is Google's specialized heavyweight reasoning model.

So you get three gears, quick answers when speed matters, balanced performance for everyday tasks, a near deep think level reasoning for the hardest problems you throw at it. The model also comes with better token efficiency, meaning it gets to the answer using fewer tokens, and Google improved its factual consistency, less hallucination, more reliable outputs. That matters a lot when you're building something real. Now, let me tell you where you can actually access this. Gemini 3.1 Pro is available in Google AI Studio right now for free. No subscription required. Just head to Google AI Studio and it's there. If you want to use it inside Anti-Gravity, download it at anti-gravity.google. Google also free for individuals. For the Gemini app, 3.1 Pro is rolling out for AI Pro and Ultra Plan users and it's also in Notebook LM for those same plans. One thing that makes anti-gravity different from other agentic tools is how it handles trust. You give the agent a task and then you have to scroll through a wall of raw tool calls to figure out what it actually did. It's tedious. It breaks

the flow. Anti-gravity solves this with what they call artifacts. As the agent works, it produces tangible deliverables, task lists, implementation plans, screenshots, browser recordings. You can review the agents reasoning at a glance, and if something looks wrong, you can leave feedback directly on the artifact, like commenting on a Google doc, and the agent incorporates that input without stopping its work. On top of that, anti-gravity isn't locked to just Gemini. You can also choose Anthropics Claude Sonet 4.5 and OpenAI's openw weight models within the same session. So you have genuine flexibility on which model handles which task. And now with 3.1 Pro as the default Gemini option, that default just got dramatically better. The speed at which things are moving in this space is real. Google went from Gemini 3 Pro to 3.1 Pro in just a few months and the improvements are measurable in the exact tasks that Aentic platforms rely on. What I'd

encourage you to do right now, go to Google AI Studio, try Gemini 3.1 Pro, ask it something complex, then download anti-gravity and give it one real task you'd normally do yourself, not just to see what it can do.