Forgot your password?
Close
typodupeerror
AI Google

Google Releases Gemini 3 Flash, Promising Improved Intelligence and Efficiency 24

Posted by BeauHD from the near-Pro-level-performance dept.
An anonymous reader quotes a report from Ars Technica: Google began its transition to Gemini 3 a few weeks ago with the launch of the Pro model, and the arrival of Gemini 3 Flash kicks it into high gear. The new, faster Gemini 3 model is coming to the Gemini app and search, and developers will be able to access it immediately via the Gemini API, Vertex AI, AI Studio, and Antigravity. Google's bigger gen AI model is also picking up steam, with both Gemini 3 Pro and its image component (Nano Banana Pro) expanding in search.

This may come as a shock, but Google says Gemini 3 Flash is faster and more capable than its previous base model. As usual, Google has a raft of benchmark numbers that show modest improvements for the new model. It bests the old 2.5 Flash in basic academic and reasoning tests like GPQA Diamond and MMMU Pro (where it even beats 3 Pro). It gets a larger boost in Humanity's Last Exam (HLE), which tests advanced domain-specific knowledge. Gemini 3 Flash has tripled the old models' score in HLE, landing at 33.7 percent without tool use. That's just a few points behind the Gemini 3 Pro model. Gemini 3 Flash has been been significantly improved in terms of factual accuracy, scoring 68.7% on Simple QA Verified, which is up from 28.1% in the previous model. It's also designed as a high-efficiency model that's suitable for real-time and high-volume workloads.

According to Google, Gemini 3 Flash is now the default model for AI Mode in Google Search.
This discussion has been archived. No new comments can be posted.

Google Releases Gemini 3 Flash, Promising Improved Intelligence and Efficiency More

Google Releases Gemini 3 Flash, Promising Improved Intelligence and Efficiency

Comments Filter:

  • Why? (Score:2)

    by ebunga ( 95613 )

    Who wanted this?

    • Re: Why? (Score:3, Informative)

      by retchdog ( 1319261 )

      Our masters, obviously.

      • Re: Why? (Score:4, Insightful)

        by sit1963nz ( 934837 ) on Wednesday December 17, 2025 @05:57PM (#65865239)
        Got to get as much money as they can stashed away before the bubble collapses and all the last in rubes lose everything.

        • How come there hasn't been an advertising bubble? Is today's advertising really efficient? Is efficiency being annoying?

          • advertising is NOT asset intensive the way AI is.
            The price for each impression has fallen over the years, the effectiveness of adverts is falling.
            As for me, I just assume that 100% of adverts are a scam and treat them all that way.. But I also run Ad blockers etc too.

            • How come google keeps serving me the same ads I repeatedly give feedback that I don't want to see? Might AI be less hallucinatory than the current ad serving software?

              • Because they are paid too do so, your input is irrelevant.

                • Csn you see how your unprincipled defense of advertising while castigating AI might seem like a huge double standard?

                  • Not at all, how much does an AI centre cost to set up, run, maintain, etc.
                    And this far NO ONE has made any money, they are racing as fast as they can looking for an end that is not yet defined

                    AI is just like the dot-com bubble, and its going to pop and leave hundreds of billions of loses, pointless hardware, etc.
                    Advertising "impressions" is just you have seen it, click though are another form of profit.
                    Advertising is also profitable, this is how google, Facebook, etc etc etc make their billions

                    I
    • Too many AI entrants; most will fail. Success depends on having a critical mass of users. So, specs-manship to attract users. BTW, other factors are required (for instance acquiring high-quality/novel training data) but the most important are number of users and engagement metrics.

  • I'm Eager For It (Score:3)

    by SlashbotAgent ( 6477336 ) on Wednesday December 17, 2025 @05:23PM (#65865161)

    I'm eager to have better and faster AI available to me.

    But, just this morning Gemini was misinforming me about well documented facts.

    I've yet to find an AI that could be trusted.

    • So its a clear reflection of its owners.."Should not be trusted"

    • Would Diogenes of Sinope like a word?

      "Diogenes advocated for a return to nature, the renunciation of materialism, and introduced early ideas of cosmopolitanism by proclaiming himself a "citizen of the world". His memorable encounters, including that with Alexander the Great, along with various accounts of his death, have made him a lasting symbol of philosophical defiance to established authorities and artificial values."

    • That's part of the learning curve with LLM's today, you have to practice strategies for each model to get reliable results on test cases where you know the answers you want, then apply thay same strategy to real use cases.
    • AI has some good use cases. Facts are not one of them.

      • Can you do the President next? Does society care about your quaint, idiosyncratic "facts"?

        Remember how wikipedia can hallucinate facts that then get cited by others, which wikipedia then cites as evidence for its hallucination?

  • Google is always trying to push Gemini on me. Someday when I'm careless I'm likely accidentally sign up for it. That's probably what they're hoping for, so they can use me for training data.

    If only there were a way to get them to leave me alone. (I know, dream on.)

    • If it's a google feature that is off, there will always be nags. I use the chrome browser and try to avoid using sync - but sure enough either a popup will appear every few days asking to turn it on, or it will just be on and suddenly screwing up my local bookmarks and passwords. Oh and installing old extensions that I tried a long time ago and in fact aren't even supported by manifest 3 - but still installed because sync decided to turn itself back on.

      My new android tablet had maybe 8 popups to turn on

  • Nasty stuff (Score:4, Funny)

    by RitchCraft ( 6454710 ) on Wednesday December 17, 2025 @05:35PM (#65865195)

    DC Comics got it right by naming one of their supervillains Gemini.

  • It's a trained regurgitator and that's it. There is no reasoning, nor learning, nor ability to design.

    • How come I can teach it to turn text into plain ASCII to render cleanly on this site, and after a while it doesn't need to be prompted anymore?

    • That's too simplistic.

      Yes, it's a trained regurgitator, but it's regurgitating statistics not training samples, which is why it's (LLMs) a lot more capable and useful than if it was just regurgitating training samples.

      Can they reason? Well, yes, to some degree, even if there is a lot structurally missing to be human level.

      - A base model can regurgitate reasoning as well as anything else

      - RL-based post-training heavily biases these models to regurgitate reasoning steps that will work towards a successful res

Slashdot Top Deals

Somebody's terminal is dropping bits. I found a pile of them over in the corner.

Close