What’s new, what can it do and how does it compare to GPT 5?

For a time when we won’t have to navigate websites manually, it also wants to show consumers how AI-generated content will look beautiful and be enjoyable to read, and get developers started building AI agent applications.

Australian engineer James Groeneveld, Character.AI’s chief intelligence officer, said he expects big things from Gemini, given the company’s DNA in artificial intelligence goes back decades. Noam Shazeer, Google’s vice president of engineering and co-chairman of Gemini, was previously general manager of Character.

“Having worked with Noam, he is one of, if not the, all-time greats in this field,” Groeneveld said.

“I’m looking forward to not only the results of Noam’s year of work at Google, but also where they are in five years. I think it’s going to be very, very exciting.”

Is Gemini better than ChatGPT?

Another reason why Google is motivated to make a big splash out of its new model is that the latest versions of OpenAI have been pretty poorly received. Dan Petrovic, an academic and consultant on SEO and generative AI, said Google’s size, expertise and vast trove of search data gives it a huge advantage, but the Gemini 3 Pro will likely be a more expensive model to run.

“When GPT 5 came out, it suddenly felt like a dumber model. If you turn off search grounding, it’s not very capable. Instant responses are really crap. Compare that to Gemini; if you turn off search and there’s no external source of information, it’s still very knowledgeable,” he said.

“From what we see in the comparisons, [Gemini 3 Pro] It’s not just slightly stronger than GPT 5.1, it’s much more powerful. But Google would use higher level thinking to produce this output, resulting in wasted time and calculations.”

Although Google said it would roll out Gemini 3 to all of its products, Petrovic speculated that it would either use a variety of models for consumer-facing products or use a faster version because the entire Gemini 3 Pro model would burn money on every web search.

Google’s announcement focuses heavily on empowering developers and content creators not only with interlaced coding tools but also with ways to build AI tool applications. Although Gemini 3 has essentially the same knowledge base as 2.5 (due out January 2025), it has been retrained and can access Google Search by touch; This, combined with its visual capabilities, may make it superior to GPT.

Loading

But given that OpenAI is focusing more on consumer-level features like chat personalities, it remains to be seen how casual users will feel about this difference.

Google search with AI Mode

To combat the threat of others using AI to eliminate the need for Search, Google’s AI Mode is a tool that takes your question, quickly performs multiple searches, and synthesizes the answers into the transcript. Google has said that it has already pushed the Gemini 3 into AI Mode, but there’s no real way to confirm this. Using my Pro account I asked AI Mode if it was using 2.5 or 3 and even it didn’t know.

Regardless, I avoid using AI Mode instead of traditional search because I want to continue rewarding providers with good information with my eyes. But I reviewed my search history from the last few days to find some of the questions I had asked, and imported them into AI Mode to see how Gemini handled them.

On a question about stopping my cats from knocking over their automatic feeders to eat overnight, regular searching led me to Reddit, where I found some good advice. AI Mode evaluated multiple Facebook, Quora, and Reddit posts to provide three answers (secure the feeder, reinforce the cover, address the behavior) with recommendations for each.

In another question, I wanted to learn more about a character in a comic book I read to my children. Regular searching led me to others asking the same questions, but AI Mode found a discussion with the comic book writer that was exactly what I needed.

The system is less successful when you need very up-to-date information. He turned down a request for what was showing in movies, essentially listing theaters near where he thought I lived (wrong) and told me to check them out.

Overall, the system provided the advantage of making multiple searches in a very short time. The biggest problem is that if everyone starts using AI Mode, eventually Quora answers won’t have new Reddit threads to feed into. Only AI-generated garbage is created to delete recent monetized clicks.

“The open web is at risk, content production is also at risk, and people are not motivated to produce more content. At the same time, more artificial intelligence trends are being entered into the indices,” Petrovic said.

“Radio pays royalties to the artist every time they play a song. Google and OpenAI should start doing the same.”

What can the Gemini app do right now?

For consumers, if you want AI to solve any problem, the Gemini app or web interface is the place to go. There is a “quick” setting using a simpler model and a “thinking” setting using the Gemini 3 Pro. Free users can use “think” only a few times a day. I set three missions for Gemini 3 to test.

First, I asked him to create a slideshow presentation comparing six dedicated mirrorless cameras. They are all a few years old and look more or less similar. It took Gemini about 90 seconds to render the slides, and despite the odd choice of cover photo with green German text, they appeared to meet my specifications. At first glance.

Loading

Each of the six had their own slide with a price range, pros and cons, and a photo. But the AI ​​seemed to have confused the Olympus OM-D E-M5 Mark III (which I wanted) with the Olympus Tough TG-5, which is a completely different type of camera. He gave a made-up price and details.

He also chose to include the Fujifilm X-T30 II instead of the older model I wanted. When he offered vague opinions (“autofocus tracking can be clumsy”), there was no way to find the reviews he sourced to learn more. Some of the graphs and comparisons in the later slides contained good information showing that this could be a great way to learn. But the system kept making up facts about the E-M5/TG-5 hybrid it envisioned.

I then asked Gemini to plan a trip to Osaka. I told them where and for how long I would be staying, while also choosing vegetarian meals and staying away from tourist traps. Again, at first glance it did an excellent job; While it got me some of the results I would get from a Google search, it also turned up some fun suggestions I’d never heard of before. But since I’m more familiar with the cameras than I was with Osaka, it’s harder to spot potential errors. I definitely double check all travel times.

Loading

The AI ​​chose four themes and embellished them throughout the four-day plan, but also focused on a specific theme for each day. This included offbeat attractions a little further out of town, such as a hike to Minoo Falls and a detour to an old insect museum, as well as less popular alternatives to the obvious options, such as seeing Kishiwada Castle instead of Osaka Castle. She also found interesting vegan or vegetarian restaurant options for lunch and dinner each day near the attractions she recommended. Some extreme details (“the owner speaks English”) still made me wonder where the facts came from.

Finally, I wanted garden design and this was where it was the most difficult. I uploaded a drawn description of the area, identified preferred plant species, and requested a visual representation, but this fell back on classic chatty and detailed AI writing. He selected four plants and gave rationales along with a step-by-step plan, but his visual representation was just a plain text drawing of short lines and letters. His drawing was also inconsistent with the planting advice.

I asked if he could provide a visual representation of the plants and he sent me some images from Getty and Shutterstock. I asked if he could create an image of what the finished garden would look like, and he enthusiastically said he could, but then a big black nothingness appeared.

Get news and reviews on tech, gadgets and games in our Tech newsletter every Friday. Sign up here.