Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Gemini 3

blog.google

Post reply on HN
Print view View original

881–890 of 1001 posts

Re: Gemini 3

#881

Post by KingMob »

dboreham wrote: post #230
mccoyb wrote: post #138

I truly do not understand what plan to use so I can use this model for longer than ~2 minutes. Using Anthropic or OpenAI's models are incredibly straightforward -- pay us per month, here's the button you press, great. Where do I go for this for these Google models?

Also Google discontinues everything in short order, so personally I'm waiting until they haven't discontinued this for, say 6 months, before wasting time evaluating it.

Generally a good idea with Google, but if the pace of model releases keeps up, nobody will be running 6-month-old models from anyone.

Re: Gemini 3

#882

Post by KingMob »

dyauspitr wrote: post #286
svantana wrote: post #84

Grok got to hold the top spot of LMArena-text for all of ~24 hours, good for them [1]. With stylecontrol enabled, that is. Without stylecontrol, gemini held the fort. [1] https://lmarena.ai/leaderboard/text

Grok is heavily censored though

Is it censored... or just biased towards edge-lord MechaHitler nonsense whenever Musk feels like tinkering with the system prompt?

Re: Gemini 3

#883

Post by rubymamis »

I gave it the task to recreate StackView.qml to be feel more native on iOS and it failed - like all other models...

Prompt:

Instead of the current StackView, I want you to implement a new StackView that will have a similar api with the differences that:

1. It automatically handles swiping to the previous page/item. If not mirrored, it should detect swiping from the left edge, if mirrored it should detect from the right edge. It's important that swiping will be responsive - that is, that the previous item will be seen under the current item when swiping - the same way it's being handled on iOS applications. You should also add to the api the option for the swipe to be detected not just from the edge, but from anywhere on the item, with the same behavior. If swiping is released from x% of current item not in view anymore than we should animate and move to the previous item. If it's a small percentage we should animate the current page to get back to its place as nothing happened. 2. The current page transitions are horrible and look nothing like native iOS transitions. Please make the transitions feel the same.

Re: Gemini 3

#884

Post by bluecalm »

I've asked it (thinking 3) about the difference between Plus and Pro plans. First it thought I am asking for comparison between Gemini and ChatGPT as it claimed there is no "Plus" plan on Gemini. After I insisted I am on this very plan right now it apologized and told me it in fact exists. Then it told me the difference is that I got access to newer models with the Pro subscription. That is despite Google's own plan comparison page showing I get access to the Gemini 3 on both plans.

It also told me that on Plus I am most likely using "Flash" model. There is no "Flash" model in the dropdown to choose from. There is only "Fast" and "Thinking". It then told me "Fast" is just renamed Flash and it likely uses Gemini 2.5. On the product comparison page there is nothing about 2.5, it only mentions version 3 for both Plus and Pro plans. Of course on the dropdown menu it's impossible to see which model it is really using.

How can a normal person understand their products when their own super advanced thinking/reasoning model that took months to train on world's most advanced hardware can't?

It's amazing to me they don't see it as an epic failure in communication and marketing.

Re: Gemini 3

#885

Post by tkgally »

hbbio wrote: post #847
tkgally wrote: post #721

Earlier quoted context omitted.

I updated my benchmark of 30 pelican-bicycle alternatives that I posted here a couple of weeks ago: https://gally.net/temp/20251107pelican-alternatives/index.ht... There seem to be one or two parsing errors. I'll fix those later.

You should add ChatGPT. I tried the first one and 5 Pro gives this: https://imgur.com/a/EhYroCE

Thanks for the suggestion. I’m not sure why I didn’t include an OpenAI model in my first round. Here’s the updated page with GPT-5.1 results added:

https://gally.net/temp/20251107pelican-alternatives/index.ht...

As your example shows, GPT-5 Pro would probably be better that GPT-5.1, but the tokens are over ten times more expensive and I didn’t feel like paying for them.

Re: Gemini 3

#886

Post by coldtea »

panarky wrote: post #697
lairv wrote: post #293

Out of curiosity, I gave it the latest project euler problem published on 11/16/2025, very likely out of the training data Gemini thought for 5m10s before giving me a python snippet that produced the correct answer. The leaderboard says that the 3 fastest human to solve this problem took 14min, 20min and 1h14min respectively Even thought I expect this sort of problem to very much be in the distribution of what the mo…

[flagged]

>This is not an incremental advance. It is a step change. This indicates a new discovery, not just more data or more compute.

To succeed this well in math, you can't just do better probabilistic generation, you need verifiable search.

You need to verify what you're doing, detect when you make a mistake, and backtrack to try a different approach.

Loos like AI slop

Re: Gemini 3

#887

Post by jacky2wong »

What I loved about this release was that it was hyped up by a polymarket leak with insider trading - NOT with nonsensical feel the AGI hype. Great model that's pushed the frontier of spatial reasoning by a long shot.

Re: Gemini 3

#888

Post by lionkor »

tails4e wrote: post #694
sedatk wrote: post #492

Earlier quoted context omitted.

Just to clarify the context for future readers: the latest problem at the moment is #970: https://projecteuler.net/problem=970

I just had chatgpt explain that problem to me (I was unfamiliar with the mathematical background). It showed how to solve closed form answers for H(2) and H(3) and then numerical solutions using RK4 for higher values. Truly impressive, and it explained the derivations beautifully. There are few maths experts I've encountered who could have hand-held me through it as good.

Was the explanation correct?

Re: Gemini 3

#889

Post by calmoo »

panarky wrote: post #697
lairv wrote: post #293

Out of curiosity, I gave it the latest project euler problem published on 11/16/2025, very likely out of the training data Gemini thought for 5m10s before giving me a python snippet that produced the correct answer. The leaderboard says that the 3 fastest human to solve this problem took 14min, 20min and 1h14min respectively Even thought I expect this sort of problem to very much be in the distribution of what the mo…

[flagged]

Your comment is AI generated

Re: Gemini 3

#890

Post by mlrtime »

windexh8er wrote: post #740
TechDebtDevin wrote: post #691

Earlier quoted context omitted.

Why is this sad. You should bw rooting for these LLMs to be as bad as possible..

If we've learned anything so far it's that the parlor tricks of one-shot efficacy only gets you so far. Drill into anything relatively complex with a few hundred thousand tokens of context and the models all start to fall apart roughly the same. Even when I've used Sonnet 4.5 with 1M token context the model starts to flake out and get confused with a codebase of less than 10k LoC. Everyone seems to keep claiming thes…

>documented problem in Home Assistant this evening. All it would take is 3-5 lines of YAML. The model failed miserably. I think we're all still safe.

This is mostly because HA changes so frequently and the documentation is sparse. To get around this and increase my correction rate, I give it access to the source code of the same version I'm running. Then instructions in CLAUDE.md on where to find source and it must use source code.

This fixes 99% of my issues.

Post reply on HN