Google Launches More Intelligent Gemini 3 Model

[MacRumors]

Google today introduced Gemini 3 Pro, its newest and most intelligent AI model. Google says that Gemini offers state-of-the-art reasoning, able to understand depth and nuance. It is also better at understanding the context and intent behind a request for more relevant answers.

According to Google, Gemini 3 Pro is the best model in the world for multimodal understanding, outperforming Gemini 2.5 Pro on every major AI benchmark. Responses have been designed to be concise and direct, with less flattery. Google claims that it serves as a "true thought partner."

Gemini 3 Pro is rolling out across Google platforms. It's been incorporated into AI mode in Search for Pro and Ultra subscribers, the Gemini app (select Thinking from the model selector), AI Studio, Vertex AI, and Google Antigravity, a new agentic development platform. AI Mode in Search will use Gemini 3 to provide new generative UI experiences like immersive visual layouts and interactive tools generated on the fly.

Google AI Ultra subscribers can also use Gemini 3 with Gemini Agent as of today, with Gemini 3 able to execute multi-step workflows from start to finish.

Gemini 3 Deep Think is even more intelligent, and Google says that it can solve more complex problems than Gemini 3 Pro. Gemini 3 Deep Think Mode will be available to Google AI Ultra subscribers in the coming weeks.

As part of the Gemini 3 launch, Google redesigned the Gemini app to give it a more modern look. Google says that it's easier to start chats and find images, videos, and reports that you've created in a dedicated My Stuff folder. The shopping experience has been overhauled, incorporating product listings, comparison tables, and prices from Google's Shopping Graph.

There are new interfaces, including a visual layout that uses photos and modules, and a dynamic view that uses agentic coding capabilities to create a custom user interface in real-time suited to a query.

Article Link: Google Launches More Intelligent Gemini 3 Model

 

[jole]

Hello new Siri!

 

[Macalway]

It's built in everywhere already, as of today. I just checked. Very cool stuff.

As far as it taking 'her' place; can't happen soon enough

 

[segfaultdotorg]

Now we know why Apple was so eager to license it to replace Siri and Apple Intelligence!

 

[mikekow80]

I want to be supportive, but till I get the early access and my home displays to have Gemini, I don't care. My parents got it like an hour after we both signed up, like 3 weeks ago. If I have to pull my phone out, I'll just keep using ChatGPT since it's got a convenient widget on my main screen.

 

[Macalway]

Now we know why Apple was so eager to license it to replace Siri and Apple Intelligence!

Maybe, timing lines up, but with hype as a factor, who knows.

 

[mxrider88]

Finally the competition is catching up with Siri!!

 

[LightProtector]

So far, it’s been better than 2.5 Pro for me. I’ve been waiting for this update!

 

[Jakesky123]

Been trying it all day and seems maybe 5% better overall, slow as molasses when using the "thinking" mode though. Still no real uses for me other than an alternate to googling things.

 

[diewish2007]

More intelligent = lies faster and more convincingly?

 

[klasma]

Apple probably made a deal for the old one.

 

[sw1tcher]

According to Google, Gemini 3 Pro is the best model in the world for multimodal understanding, outperforming Gemini 2.5 Pro on every major AI benchmark.

Meh. I'll wait for Gemini 3 Pro Max (Ultra?) which will be the bestest model in the world for multimodal understanding.

 

[Kylo83]

And we have image playground lol

 

[JSRinUK]

I’ve just embarked on an “AI Challenge” - challenging various AIs to create a bug-free coding project.

I saw that Gemini had a “Thinking 3 Pro” option in the drop down, so I gave it the challenge.

It’s early days yet - I have to test out all the code from all the models.

However, I was curious to see Gemini’s thinking when it said:

While the user's mention of "Qwen3" seems slightly off, I'm adapting to their intent, leaning toward a Qwen2.5 variant.

What current model doesn’t know about Qwen3? And why didn’t Google Gemini just, you know, search the web? If it didn’t recognise it, couldn’t it just “Google it”?

That doesn’t bode well.

 

[SKeijmel]

LoL so apple bought the old model. We getting next year in "siri"

 

[Apple MessagePad]

These press releases are so dumb. As if they'd launch a less intelligent system. It says the same thing their past releases said.

It's much like the stupid post MR makes every time there's a new Safari Technology Preview. They go with a headline that says bug fixes and performance enhancements. That's literally the entire purpose of it. Might as well save yourself the time and just say "Update has updates.", as if that's actual news.

 

[ipedro]

I've been "forced" to use Gemini at work because we have Google Workspace accounts, but I've been secretly using my chatGPT account instead because it's been miles ahead. Over the last few weeks, Gemini seems to have taken leaps forward, even with the same old 2.5 Pro model. From using 3.0 today, I'd say it's on par with what I expect from chatGPT and I might finally stop using openAI's model.

The prospect of losing this war was an existential crisis for Google. They seem to have taken that seriously and worked to climb back up. Tim Cook and his VPs missed it entirely. There's still time for the players to establish themselves but Apple really has to deliver early next year.

 

[stars_fan]

It’s meh

 

[Surrylic]

I've been "forced" to use Gemini at work because we have Google Workspace accounts, but I've been secretly using my chatGPT account instead because it's been miles ahead. Over the last few weeks, Gemini seems to have taken leaps forward, even with the same old 2.5 Pro model. From using 3.0 today, I'd say it's on par with what I expect from chatGPT and I might finally stop using openAI's model.

The prospect of losing this war was an existential crisis for Google. They seem to have taken that seriously and worked to climb back up. Tim Cook and his VPs missed it entirely. There's still time for the players to establish themselves but Apple really has to deliver early next year.

They appear to be making the easy and right choice by just leveraging Gemini, at least. Can’t wait for that update. Even the older versions of these are 1000x better at answering anything than Siri is.

 

[jdawgnoonan]

These things are not “Intelligent”. They pattern match on language.

 

[Boeingfan]

Now we know why Apple was so eager to license it to replace Siri and Apple Intelligence!

I just hope that Apple doesn't choose Gemini v1.0 to replace Siri, though, having said that it may still be a step up.

I asked Siri to shuffle my 90's favourites playlist on my iPhone (I don't use streaming) a couple of days ago, and she replied, "To do that, you'll need to be online." 🤷‍♂️

 

[novagamer]

Although I loathe Google I'm going to get Ultra once Deepthink 3 comes out in a few weeks, I have a specific use case and need to go through tens of thousand off lines of code and Gemini's context is unmatched (Sonnet 4.5 has already done a pass, but I need another one).

This stuff is pretty useful for specific things. Really, really useful.

Glad Google caught up, hopefully Anthropic gets the new Opus out soon, and especially that they update the training date to match this model.

 

[pdaholic]

I've been comparing Siri and Gemini while helping my son with his homework. Siri is faster at simple tasks, but Gemini is amazing at solving complex problems like word problems. Siri doesn't come close.

 

[Kkspire]

I've been "forced" to use Gemini at work because we have Google Workspace accounts, but I've been secretly using my chatGPT account instead because it's been miles ahead. Over the last few weeks, Gemini seems to have taken leaps forward, even with the same old 2.5 Pro model. From using 3.0 today, I'd say it's on par with what I expect from chatGPT and I might finally stop using openAI's model.

The prospect of losing this war was an existential crisis for Google. They seem to have taken that seriously and worked to climb back up. Tim Cook and his VPs missed it entirely. There's still time for the players to establish themselves but Apple really has to deliver early next year.

Apple didn’t miss out at all. They didn’t waste 3 billion dollars scraping people’s personal data. It’s a high money race with almost no returns. Apple ain’t stupid. And they aren’t google. Their model is to not collect data, not to collect as much as they can and sell it.

 

[dcingie]

Objectively better than Grok under the benchmarks?