Announcing Gemini 3.1 Pro! Google's smarter model for your most complex tasks just shipped 🚀 Demos:

Gemini 3.1 Pro can generate animated SVGs that are web-ready from simple text prompts too! They stay sharp at scale and are smaller than standard videos.

Gemini 3 Pro vs. 3.1 Pro

Gemini 3.1 Pro is Google's new baseline model for complex problem-solving. This model has a big leap in core reasoning. It scores a 77.1% score on the ARC-AGI-2 benchmark - which tests the ability to solve brand-new logic patterns. That's more than double the performance of 3 Pro.

This is awesome. Hoping it is a big step up from Gemini 3 in antigravity

Rolling out to Antigravity in the coming hours. Do let us know if you have feedback!

I will for sure!

new model, devs be like

The hype factor is 😂

it's funny cause it's true 😁

the pace of model releases right now is insane. been using gemini for coding tasks and the jump between versions is genuinely noticeable each time

Glad to hear it's ben your go-to for coding. A lot more we are planning to improve here for fullstack, gaming and more.

Yo, this is cool stuff, I guess I am seeing the new frontier for LLMs to build complex UIs and user-facing interfaces, I hope we can that 3.1 "sees" the visual stuff rather just crunching tokens. Excited to try it out over when it is available for general use.

For sure! Using it for a lot of UI work atm. The Gemini architecture is natively multimodal. So the model does possess the ability to "see" and process visual input directly to inform how they generate code.

Nice, with skills it will be even better

Absolutely. Been finding the same with a few frontend + performance related skills of late.

Version 3 was my go-to for UI design, but if this one is even better, it's a big win.

Happy to hear and I can say 3.1 Pro is even stronger on front-end UI!

Can't wait to try that new model in Intent

go Gemini!

🚀🚀🚀

damn, the world is really moving so fast. Let's have it

🙏 and more to come!

Hope it gives better results

That's been my experience. Hope that you have a good one with the model as well

Thanks man for even responding you have been a role model🤩

This is just incredible on how it can buid these crazy generative Ui interfaces! It does poke your imagination a bit

Totally does. As someone that took a while to learn @threejs + shaders really enjoying how much Gemini has lowered the bar for building scenes + UI out.

@threejs Time to build that animated fish tank for my daughter then! 😉 🐠

🙏🙏🙏

This is amazing, congrats!

Go team! 🙏

Gemini 3.1 Pro quietly moved the frontier while most were still debating 2.0. Progress doesn’t ask for permission — it just ships. Well played, Google.

The most important thing i see with this release is the upgrade with UI designs and animations Ngl, the gemini models have gotten far better with creating design and animation Using gemini 3.1 pro makes everything a blast, I don't feel stress no more

Glad to hear that design and animation have felt stronger. We'll be continuing to work on getting stronger here as well as getting better at fullstack web.

been testing 3.1 pro all morning and the structured output is noticeably better. context window handling feels way more reliable too

Fantastic to hear on both fronts. As you try it out if there are any gaps, please feel free to let us know!

Smarter models at the same price point is the pattern now. Curious how this affects the Gemini vs Claude vs GPT decision for builders running production workloads.

Biased, but I think Gemini offers pretty competitive quality for the cost. Always happy to hear feedback if there are gaps here for builders :)

@thebuildrweekly Gemini is the best overall value for consumer plans. It just cranks all month and barely hits limits. It does have serious peaks and valleys on inference though. Hopefully 3.1 smooth that out.

@thebuildrweekly Validating to hear this and I think 3.1 will "feel" more smooth. If you come across gaps, do hit us up!

Keep going !

🚀🚀🚀

@addyosmani when is going to be available for coding tools ? (Antigravity, cli, etc)

Available in preview via the Gemini API in Google AI Studio, Gemini CLI, Android Studio and Google Antigravity.

Sorry I meant to say available using the Google AI pro plan

Rollout has started for Gemini app and NotebookLM users on Pro and Ultra. You should have it soon.

Ty !

Bring back the old Ai Studio limits please

Let us use an oauth sub in opencode and then we will use it!

Animated SVGs from LLMs sound promising, but consider browser rendering bottlenecks—complex animations quickly hit CPU limits, negating size benefits versus optimized video codecs.

Gemini 3.1 Pro’s “smarter” claim demands rigorous benchmarking against open models like Mixtral 8x22B—specifically on reasoning tasks exceeding simple instruction following.

Nice! What environment or CLI or IDE or TUI etc are you using it in right now?

Tasks are just window dressing-real work needs rhythm.

Great insights. Have to play around some video generating ideas 💡

Whats the word on a new Nano Banana model dropping soon? We need cheaper image generations lol

Will this finally replace copywriters who know nuance?

Awesome. The key is whether this translates into cleaner PR diffs and fewer review roundtrips. Model capability matters, but SDLC outcomes are the real scoreboard.

A coding gem is now the perfect environment to work with my prompt constructing tool CWC :)

The reasoning jump in 3.1 Pro looks massive! As a student, I'm most excited for the agentic coding part the idea of a model actually 'verifying' code instead of just generating it is a total game changer for my workflow. Can't wait to test the demos! Congrats team.

Super exciting and just in time for my new project. Let’s go.

The ARC-AGI-2 score is interesting, but the decision teams actually face is simpler: which model gives consistent, debuggable outputs for the task we need to automate reliably. Benchmark improvements matter less than behavior stability in production workflows. Curious how 3.1 holds up on that.

Generative AI is rapidly maturing beyond simple task completion. The leap from 3 Pro to 3.1 Pro signals a shift towards models anticipating user needs—invest now in prompt engineering skills to unlock exponentially greater value as these.