Gemini 4 Argon | Hacker News Reader

Gemini 4 Argon

deepmind.google

1,439 points737max1,027 comments

See also: Gemini 4 Argon for Robotics - https://news.ycombinator.com/item?id=50074869

We evaluated Gemini 4 Argon in 100 unsaturated, competitive and cooperative coding environments. It's roughly in between Fable 5 and Fable 5.1 in our testing, so it's maybe 2 months behind the frontier.

Google's major releases have usually done pretty well on our benchmarks because they're good at hard problems and open-ended decision making (with a well specified prompt), which obviously doesn't always translate to user experience. In particular, older Google models struggled to work efficiently in a harness. They seem to have solved that with Argon.

We did see a new problem which I hope they can fix. This is the first near-frontier model we've caught in infinite reasoning loops in a long time. We also saw a few examples of it failing to properly delineate reasoning blocks, so it has a higher failure rate than its competitors.

My overall impression is that Argon is a near-frontier model for engineering design, code analysis, and chat. It definitely has the intelligence to be a competitive coding assistant product, but a good coding assistant needs to be polished, so I hope they can clean up the CoT bugs.

Data at https://gertlabs.com/rankings

At this point I'm interested less in how well a model can code, but more in how well it can produce code that humans can understand.

> This is the first near-frontier model we've caught in infinite reasoning loops in a long time

Depends on where you draw the line for "near frontier model", but the current leading open-weights model, MiMo-v2.6, has exactly this problem. They released a fix on Hugging Face[1], but as of two days ago the model available via Xiaomi's own API still sticks in reasoning loops.

[1] https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-MOPD

I wonder if it's as good at "soft skills" as previous Gemini models were. Google was typically better in that department in my experience than either OpenAI or Anthropic, whose frontier models are focused on coding and verifiable task at the expense of all else.

Thanks for this substantive review -- anecdotally, I've felt many in the gemini series have pretty good visual sense/image processing -- do you guys have any tests in these areas, and if so, what is your take?

>> This is the first near-frontier model we've caught in infinite reasoning loops in a long time

This seems to be such a Google thing. Anyone who has used Gemini for long enough has encountered these 'She loves me, She loves me not' spam loops

> failing to properly delineate reasoning blocks I also experienced this

The second prompt I tried with Argon got stuck in a loop. That’s not encouraging.

What programming languages? HTML/Javascript or C++/Rust, it makes a difference you know.

That's been an issue since Flash 3.5, they probably have been unable to pin down the cause

What harness do you use for this?

I haven't tried it for coding yet, but I ran a couple of my recent curiosity driven queries from Claude (mostly used Fable and Opus 5.5) through it.

This is a clear step change, I'm genuinely amazed.

Its answers are to-the-point, logically structured, clearly written. Like a really well made mini essay on your question.

It also really likes to build little simulations / models that you can tweak the parameters on. Amazing to explore how different assumptions affect economic scenarios, for example.

It's also great at picking (sourced) illustrations.

I'll be using this for everything that isn't coding from now on.

It’s wild how good it is at explaining nuanced and complex health topics.

As an example, I recently had an OCT scan of my retina because I'm a “glaucoma suspect.” I'm a glaucoma suspect because I have an unusually high cup-to-disc ratio, which is a pattern typically seen in glaucomatous patients.

As part of this OCT scan, you get a report back on your RNFL, which is basically a report on the health and thickness of your optic nerve(s) and retina(s). (Dramatically oversimplifying, of course.)

My ophthalmologist struggled to explain my OCT scan results to me and instead sort of shrugged and said, "It's probably because you're a high myope, but we'll keep you coming in once a year.”

But my scans look horrifically bright red — meaning I'm in less than the first percentile of retinal thickness for my age. Obviously concerning enough to do my own research…

ChatGPT 6 Pro and Opus 5.5 have struggled to make any sense of the clinical implications of these scans. In contrast, Argon was the first model to:

1. explain how the OCT machine works

2. detail the physiological mechanism that causes the cupping and thinner RNFL layers in my scenario

3. sum it up into a clinical recommendation (similar to what my ophthalmologist recommended, with the addition of basic supplementation and lifestyle modification changes)

4. provide an interactive visualization of the topography of my eye vs a glaucomatic eye: https://optic-nerve-3d.netlify.app/

I'm blown away.

Totally agree. This is my new favorite general-purpose/non-coding model. It's surprising how good this thing is... And fast, too.

Google's models have usually been pretty good at general query stuff, they fall flat when doing harder tasks.

> I'll be using this for everything that isn't coding from now on.

why? is it not so good on that? or is there something better performance, or better price for same performance?

"curiosity driven queries" – love this phrase, i'm def going to start using it :)

My first reaction after reading the announcement post is that the coding scores look like a weak point.

Big if true. Opus 5.5 is a lot better than 5.0 but I find it maddening thay no single model released in the past ~year can write prose with a clear and coherent voice. I blame RL on programming tasks.

I’m curious about your curiosity driven queries. Mind to share?

With 3.8 Flash on High I noticed a huge upgrade in intricate discussions related to physics, material properties, real-world application, and so on. Things that I've had a hard time getting LLMs to be able to discuss accurately. They often discuss textbook knowledge fine but can't apply it properly or can't think outside the box at all. 3.8 High understands applied material sciences in a way that GPT Astra flunks at often. I asked Gemini to explain that result, and it said basically that GPT was merely book smart in a classical sense and that it was likely undertrained on more obscure material.

I have had the google search results ai also do these simulations and code generations unprompted, I thought that feature already existed.

I have a project I am working on where I wanted to have nice docs, and so I have some slop docs sitting there as placeholders until I get around to writing them, awful claudish stuff

I've actually spent some effort trying to make them halfway decent placeholders by putting together a skill based on diataxis framework (https://diataxis.fr/) that at least got everything into the right location

just took a pass at rewriting them w/ argon and am very pleased, big improvement

Agreed, its communication abilities are very good. I'm a big fan of Opus 5.5 but Argon just flat out writes better.

I've found it on par with Opus 5.5, what don't you like about it for coding?

Just praying they don’t nerf it

We have this same point made every release. I wonder if models actually are getting better or if these companies are just secretly hamstringing the previous models before releasing a similarly performing model

I'm using this model via API. I am most excited about its ability to write the english language (or any language). My least favorite trend on twitter has been people posting 2D illustrations/3D renders as a proxy for model capability. Well-written prose is far less likely to go viral but far outstrips the importance rendering a fantasy world in one shot.

I've been trying it out via the API and the model has been a delight so far. Clear in the head, not RL-fried, distinct writing style, and big model smell. That said, the em-dash is back w/ this model.

That said, I would be so frustrated with Sundar Pichai if I was on the Google team. The botched rollout of this model seems to be entirely a matter of capacity, and Pichai has been more than happy to give it to anybody but his own team.

I've been using it for a while internally at Google. I prefer Argon over Opus 5.5 because of the clear thinking and directness. It phrases things in an organized, straightforward way.

Em dashes are underappreciated — they are visually pleasing, not cluttery like semicolons or definitive like colons.

can confirm this. I didn't have high expectations for AI writing good long-form content in Indonesian. However, I recently tested it on some of my Indonesian writings and asked Argo to rewrite them while maintaining the original tone—and it absolutely nailed it.

I missed the em dash — ngl

> Well-written prose is far less likely to go viral but far outstrips the importance rendering a fantasy world in one shot.

I'm not entirely sure you're correct. Your claim boils down to "Writing natural human prose" vs "Broad use of complex software and spatial awareness". My gut feeling is that there's more untapped and fruitful applications in the second group than the first - at least in the short term.

I was kinda hoping I'd get to the end and it would say, "In fact, this entire post was drafted by Argon!"

What was botched about it?

Unpopular opinion: em-dash is fine

I added Gemini 4 Argon to our video creation benchmarks.

The results are a bit disappointing given Googles scale and training data.

The results are on par with cheaper Chinese open source models to be frank. https://cliphou.se/benchmark/

> That said, the em-dash is back w/ this model.

noooooooo

> far outstrips the importance rendering a fantasy world in one shot.

Given FelonyBench, do we also have a FantasyBench?

People should appreciate Google more for how all of their models have:

- Natively multi-modality

- Enormous context windows

- Reliably cheap prices

It's like they don't release a model until it passes the above checklist.

Most providers have some mixture of these three occasionally, but with Google, every model is consistent. This kind of consistency is really nice to have.

And the most reliable. A year and a half ago this was the exact reason why I switched to Gemini API from OpenAI. OAI would make me wait up to 10 minutes for the first token in the most egregious cases. Gemini served me almost immediately every time.

Good point. On that note, it may be a bug but I noticed Argon compacting well under the 1M max in Copilot (around 400k) -- when set to 1M context. That may be a Copilot thing but I've never seen that with other models running 1M context. Could be a day-1 bug.

Outside of that, behavior is quite good, a bit slow, and definitely different than both GPTs and Claudes -- which is appreciated.

Consistently buggy you mean. 1 in 10 tasks get stuck in reasoning loops or error out.

So cheap that people who pay for pro don't get it as a coding agent in antigravity lol

>Reliably cheap prices

do you have to pay those cheap prices eventually or can you just accrue an unlimited amount of overdue charges?

Gemini 4 Argon is a new leader in Pareto-frontier. Looks like, Google is back in the race.

https://arena.ai/leaderboard/agent/chat/pareto

For almost same intelligence (+15.56%, +15.18%), drastic difference in cost/task ($3.51 vs $0.41)

Why do these benchmarks often show models running on Max i.e. Opus 5.5 (max) is on the frontier? I would think it is never really economical to run max

> Gemini 4 Argon is a new leader in Pareto-frontier. Looks like, Google is back in the race.

For chat, yes. But switch to code on the same page, and the picture changes. OpenAI takes the lead.

Muse reminded me we should have a time based chart as well - this year I hope we will begin to see time to completion stats at various intelligence tiers. The waiting makes a lot of things more tiring.

Every release I see a new benchmark with a new frontier. These graphs are just meaningless now

Image->html test for Argon. I'd put it at B tier, just under gpt 6.1 sol. It isn't A tier like Opus 5.5 or Astra.

Argon build: https://html.non.io/lcars-argon/

GPT 6.1 sol: https://html.non.io/lcars-gpt-6.1-sol/

Opus 5.5 (currently SOTA for this): https://html.non.io/lcars-opus-5.5/

Original designs: https://diffui.ai/app/canvas/5093e689-1e74-4f26-b632-2a4500f...

I don't know how you rate, but the Opus 5.5 is literally stuttering and lagging when scrolling on a Macbook Air M3 (for me).

(Chrome, also incognito)

One interesting observation - it's the only one I've seen make an ASCII representation of the layout of each of the pages before building: https://image.non.io/46e0cce3-34d9-4fcb-8775-bfc2c70bc4bb.we...

Very subjective, but I'd say the Argon one is the best here. The other two don't quite do LCARS correctly, there are a bunch of lines/curves that are the wrong angle, don't line up, etc. The Argon one looks to me the most like a designer would have done it. Of course this is up to personal taste to some extent.

Argon's build is pretty much unusable on mobile, the header takes up half the screen

Opus one has by far the worst performance, the only one that's stuttering and making my laptop fans very noisy.

Opus is way slower in firefox on my machine than the other two.

Damn opus 5.5 is just such a good model - what kind of moloch sacrifices are they doing in there

B tier? For me the argon one is really quite bad on mobile; the nav is messed up enough to make the page kinda unusable.

On firefox on linux, argon looks great and the other two have graphical issues.

That Argon build is really bad on mobile fyi

Jesus Christ, so much "stuff" crammed into so little space. What a nightmare of a website.

Off topic, but I want to customize a starship now

How on earth can you say one is better than the other? It is subjective.

Google continues to be the best in German language and Gemini 4 is a significant jump in our benchmarks

https://dach.peerbench.ai/compare?models=google%2Fgemini-4-a...

They removed the 3.1 pro model from their offer, and version 4 is locked to ultra customers, so basically pro customers are paying for gemini 3 flash.

State of affairs [1]:

> My whole project and career depended on the 3.1 pro workflow. Please, bring it back

[1] https://discuss.ai.google.dev/t/please-bring-gemini-3-1-pro-...

It's a great model. I have had it trying to decrypt the Voynich manuscript for the past couple of days without being biased by other research/other people's attempts. If anyone is interested: https://github.com/d2kx/voynich-research

How ironic that the Readme is completely unreadable, like gibberish rivalled only by the Voynich manuscript itself.

But id the progress of decoding it is legit, you should write something properly about it!

I am still lost as I am not linguistically trained well (jargon-free didn't help for me).

What kind of progress was made with Argon with the Voynich?

Have u tried this with GPT Astra as well?

Wow I actually tried this with fable a few weeks ago and got nowhere

It looks like you've managed to decode it? Is that legit?

I saw 'jargon-free' and my eyes immediately glazed over

Adding my five cents: initial usage through OpenRouter has been great. Very fun model to spar with, clever, intelligent. On par with Opus 5.5 but without the Claudism's and more of a Sol type of conversation. Though speaking of Sol, I've been very dissapointed with Sol 6.1 to the extent that I feel OpenAI has neutered it compared to a few months ago. Especially when giving it goals it will happily run in dumb circles just burning tokens without any real work.

Here are sites with ongoing measurements checking if a model has been nerfed:

https://www.bridgebench.ai/nerf-bench

https://marginlab.ai/trackers/claude-code/ https://marginlab.ai/trackers/codex/

https://github.com/ninjahawk/livenerf

https://isitnerfed.org/

Definitely agree. Sol 5.6 was great when it was released and for nearly a month.

Since then they cut the price but the intelligence (of 6 and 6.1) has gone down to Sonnet 4 levels. I am back in a world where the model is lazy, cuts corners, ignores existing coding standards and context, makes stupid mistakes etc.

I am not sure how those models still rank so high in benchmarks when real-life usage is so obviously bad.

Sol 6.1 was released two weeks ago!

I ran it through my own coding eval today using pi/OpenRouter (11 small bug fixes from my own projects). It solved 10 of 11. Compared to Sonnet 5.5 and GPT-6.1 Sol, a median run cost me $0.52 vs $0.18 for Sonnet and $0.09 for Sol, and took ~4 minutes vs 32 seconds and ~1.5 minutes.

It thinks a ton, ~24k output tokens a run vs ~3k for Sonnet. It missed a solution by thinking its way into a 10 min timeoout.

Something to look out for - if you're using pi, it doesn't have Argon in its catalog yet and reports cost at Kimi's rates, about a third of what OpenRouter actually bills.

"Argon autonomously turns raw video to a physics-accurate simulation, estimating depth and physical boundaries, and writing the exact simulation code required to replicate those real-world environments."

This is crazy if it can work accurately.

Geometry could be understood today with the model but what also really matters is the sensing, friction, mass how smooth the subjects are, dexterity is a very difficult problem so solve, not sure how simulations could handle such factors

Every animal can do this, in order to run and jump and catch.

I’m curious at the enterprise level does Anthropic have staying power. My sense is when you use it with your own custom harness, it doesn’t, I can switch between Opus 5.5 or Gemini Argon and not really care as long as capabilities are similar. This is good for the consumer, and bad for the model developers.

Hmm how does your custom harness compare to Claude’s?

I’ve tried the various alternatives (opencode, pi.dev) but for “have several physical machines, each running an orchestrator and 5-10 agents, and each agent spawning its own subagents to do work in parallel, all of that on a tanky GitHub project boards” level of work only Claude was doing work dependably.

I can have a few orchestrator leads which I talk to and brainstorm, and let them file tickets, organize them and assemble teams of agents to work of the various problems. And all of that with just a smidge of custom tooling on top of Claude Code’s default.

Plus don’t have a budget to spin all of this on cloud runners so I made sure to do as much as I can with local work on several machines I own. Was quite easy to achieve with Claude.

Do you have something similar built on top of a open source harness, or is it from scratch?

It depends on how expensive the model is to serve. If it turns out OpenAI and Anthropic can serve the models for a fraction of the price versus Google/Meta, then Argon/Muse remain a curiosity from the investor's perspective.

We won't know the answer to this until IPO and until each company reveals the economics of their models (training and serving).

Unsurprisingly, the pelicans are excellent. Google are the only lab who have publicly shared that they have specifically trained for SVG output - https://twitter.com/sunjiao123sun_/status/202455551655137292... - and they've had great pelican results in the past too (which they have shown off themselves - https://twitter.com/JeffDean/status/2024525132266688757).

Here's low, medium, high. Even low is better than pelicans from any other lab: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...

low cost 10.185 cents, medium was 11.098 cents, high was 17.166 cents.

The costs are similar because it used more output tokens than reasoning tokens for all three pelicans, even for high.

Me in 2025 https://news.ycombinator.com/item?id=44759781

> If it's on HN and is a meme at this point, it will end up in the training set.

> It's kind of fun to imagine that there is an intern in every AI company furiously trying to get nice looking svg pelicans on bicycles.

Google listened and assigned a few interns, I guess, for the task

I have a silly pet "project" that I've been using with different models, where I provide an image of a patch sheet for the Behringer Poly-D synth and ask for an html version and this is the first time I get something that, while still needing some tweaks, basically looks like the patch sheet. Before this, the closest had been Fable, but that still was nowhere near the actual patch sheet.

This is the sheet: https://fernandoipar.com/content/music/poly-d/gemini-argon/c... (the annotations on it are by me)

And this is what Argon created: https://fernandoipar.com/content/music/poly-d/gemini-argon/b...

For this simple prompt: https://fernandoipar.com/content/music/poly-d/gemini-argon/p...

I'm honestly impressed.

Recently, I’ve discovered that Claude can render great SVG versions of user interface screenshots. I ask it to substitute simple SVG friendly landscapes for any photographic images that happen to be part of the screenshots. The results are much better than they should be, especially with Opus. Usually good enough to be used on a website as a way to get an infinitely scalable, much smaller version of a screenshot. But clearly, I should try the Gemini models on this real world task. That’s where we’ll find out if all of this training to beat the benchmark actually pays off.

You can really see the effort in the details.

Subtle highlights on the metal frame, sweat beads, jaunty little hat…it has style. At least compared to the rather literal ChatGPT outputs. https://static.simonwillison.net/static/2026/gpt-pelicans-gr...

For those who haven't seen it, Argon can use fourier circles to generate an animated 3d pelican -- I think the pelican is 3D and the motion was added by an inverse kinematics solver? https://x.com/fofrAI/status/2110043023954284889

As a cyclist, I tend to pay more attention to the bike shape than most people. Everyone knows what a bike looks like, but as with LLMs, most people would struggle to draw one - correctly connecting all the different tubes at correct angles etc.

I note that even these bikes are not actually rideable. For a bike to be stable, the front wheel axle needs to extend forward a little ahead of the fork - the fork must ”bend forwards”. This is what actually keeps the bike upright, not gyroscopic effects or some such. Given LLMs are generally good at basic maths, im surprised they get this wrong.

Maybe a bit less so in this case though. As per the announcements, the model was actually trained for this prompt, so was fed some training data. I can well imagine the data there was just not so good.

Simon’s headstone is going to have a pelican on a bicycle, isn’t it

At some point Simon needs to move from Pelicans to something more complex. Every time a new model is launched by major Frontier company, I already know with certainty that first HN comment on the news would be about how Model is drawing the Pelicans.

Why are all these AI's making the pelican's pouch so large though? Unless they have some huge fish in there or something they don't tend to look like that... Is it that it's drawing a "cartoon" pelican or something?

https://en.wikipedia.org/wiki/Pelican#/media/File:Pelikan_Wa...

It's obvious OAI hillclimbs the image generation -> SVG loop. I'd put money down on that being why OAI's pelicans have a yellowish tint that matches the yellow tint of OAI's earlier image models.

It does very good at Pac Man. I would say as good as Opus 5.5, it was slightly cheaper, but took twice as long.

Argon: https://jonclegg.github.io/pacman-bakeoff/dev/#gemini-4-argo...

Opus 5.5 https://jonclegg.github.io/pacman-bakeoff/dev/#claude-opus-5...

Wow, those are I think the best looking pelicans ever so far

How many cicles before pelicans become photoreallistic SVGs and them come back to stylistic artistic abstract renderings?

Also, that first tweet contains a commentary that, for me, explain a lot about the public misconception about how current AI works:

> wait, does this mean you guys are actively training svg generation? i was mostly amazed by it thinking the models doing better at it by itself due to becoming better at reasoning.

No, it's not better reasoning that drives current AI, it's better examples we have been giving them. Whole army of people generating content for the frontier labs.

Also, reasoning improving the out of distribution part, like computer use, tool use etc.

simonw, your pelicans are so famous now that the frontier labs can just RLVR for SVG images explicitly.

It's a simple manner of prompt...generate concept art with nanobanana, measure, llm creates svg with text, measure the distance between rasterized svg embedding to nanobanana art....do post-training RL on a 3 million such images and matching embeddings.

I swear this will be a canonical example of Goodhardt's Law in action.

You have to switch to a new benchmark that isn't so easy to run a verifiable rewards pipeline on...Perhaps a longer horizon task where it's easy to mess up midway.

doesn't this mean that you should change the pelican test to something else?

It starts to be so good that the model doesn't really matter. Each model starts to have its own personality, and it's the personality of the drawing, like a real artist, that defines the style of the pelican, and that makes it harder and harder to decide if it's good or bad.

That pelican left leg stays so firmly on the left side of the bike. Superintelligence ftw

Front bicycle wheel in all Argon examples is still wrong. This bicycle will not be stable. Axis of rotation of front wheel needs to be in-front of axis of rotation of steering wheel. Frontier VLMs still do not understand how bicycles work.

You gotta fix your tool on mobile browser, both firefox and chrome don't work.

Excellent, but not perfect -- if google really is training to pelicans, then clearly they are not capable of perfect pelicans.

As one point, the high pelican's arm is see-through/messed up.

They do look worlds better than previously, to be fair.

Since we know jagged intelligence is a thing, how useful of an indicator is the pelican test, regardless of if labs are optimizing for it or not?

At least you should make sure this post is entirely generated by agent, otherwise I'll not grant my agent the credibility score on this comment.

Regarding the pelican test. Maybe someone will know this.

Modern models can make nearly photorealistic games in threejs. So why do they struggle with these pelicans?

The fish in the basket on high is a lovely touch.

The fish in the basket is a nice touch

Honestly surprising if other labs aren’t training for svg generation. Feel like SVG’s are going to be really useful for dynamic UI generation stuff that goes beyond cobbling Html together

> Google are the only lab who have publicly shared that they have specifically trained for SVG output

wasn't the whole point of pelicans is that labs *wouldn't* train for that?

so now the benchmark is ruined

still radial spokes on both wheels of the bike which is disappointing. no real bicycle is built this way

I've been dismissive of the SVG-of-pelicans benchmark, but last night ChatGPT (on the website, toggle set to GPT-6 High) generated some SVGs and included them in our chat. I'm not sure that this feature was that valuable in this case, but it seems that textually generating SVGs is now a product feature at least for ChatGPT. The SVGs had some clear issues and I found myself wishing that it had done a better job: the nose of the rocket is overlapping with the rocket body, and in the second image I think there's only supposed to be one orange line but it tried to add styling or something that went wrong.

<https://tools.simonwillison.net/svg-render#%3Csvg%20font-fam...> <https://tools.simonwillison.net/svg-render#%3Csvg%20font-fam...>

- I was having trouble embedding the SVGs as links, and ChatGPT recommended Simon Willison's SVG Renderer as a workaround.

Just my 2 cents on the pelicans benchmark: I think it might help if you were a bit more explicit about not taking this one benchmark too seriously. I know you say that at times, but at other times (such as the comment I'm replying to) you elide it which can read as a seriousness about the benchmark.

Nobody cares about the AI heckling pelicans.

Give me reverse engineering bench and agentic coding.

Leave making visuals to humans who actually have aesthetic sense unlike whatever made this mess

Holy shit, these are like art. Especially the bikes.

The models are solving Millennium Prize questions and making progress on the Riemann Hypothesis, yet the top comment is still looking at pelicans somehow. To me, it feels even more superficial and pointless than arguing about the presidential candidates' tie choices in a debate.

what is even the point of these now? labs are most definitely training on this question now. I feel like this test scratches the surface of a "real" test for these new models

One thing I've observed Gemini being the best is in the context of non-western stuff. Like, when I ask about obscure srilankan food, or chat about Indian/Tamil movies or songs, Religious customs & beliefs, Gemini almost always answers accurately and very close to how helpful they are in Anglosphere related topics. ChatGPT also does do a bit better recently, but Claude/Grok are absolute trash on those fronts.

Google's years of research project into Indic scripts, image trainings, translation efforts & transliteration efforts is literally paying off. The AI adoption from non-English, non-Chinese population will only if you have these contextual localisations, which I think Gemini is poised to do well.

For example, you can ask for a shopping list for a marriage function in a town from Tamil nadu (the marriage custom varies by districts), and Gemini provides contextual and helpful answer (like what is the custom followed there, what is gifted traditionally, etc), whereas if you ask Claude/ChatGPT, it talks about marriage registeries or pretty generalised answers which is not relevant or useful.

100% agreed. When I ask to generate an image or a poster with an indic language, chatgpt and claude always make gibberish words for something that is not at the center but gemini nails it almost every time in first attempt.

I seriously doubt their ability to double API prices on Dec 1. Token prices are in a clear downward trend and Google should maximize revenue by competing for volume. Currently OpenAI is the only US frontier lab trying to seriously compete with open-weight models that you can run yourself for an order of magnitude lower than what frontier labs want to charge

I’m running 6.1 sol as my main driver on Hermes these days. It does so much for so little cost.

The goal is to move people along. By then they'll have something as performant but cheaper for them

When you are on "gemini.google.com" there is a little lock icon next to Gemini 3 and Gemini 4. When you click it you get the prompt to subscribe (pay). Once you have successfully done so, the Gemini 4 option disappears!

Pretty misleading.

Maybe it’s just an incomplete rollout not fraud lol

Looks like safety features are a bit strange.

https://imgur.com/a/fWo92RI

Seems to be just deleting anything that a safety filter triggers on. I don't know if I like this more or less than what Anthropic does where they train the model to actively refuse.

Gemini’s system has always been like this. I used it heavily from last September through March this year, and have been using it on and off since then — this has been the case since at least Gemini Pro 3.1. And that’s not even the worst of it. When its system detects something it deems “seemingly inappropriate,” it just freaks out. What I’ve personally experienced includes: (on the web version) wiping Gemini’s previous conversation content — once; and outright disconnecting Gemini from all prior conversation history, meaning a complete memory wipe — eight times. All without any notification. My conversations were pure text, just everyday chat about daily life and how I was feeling, absolutely nothing that would count as a violation in any general sense. But the system would just randomly lose it. After eight consecutive memory wipes, I mostly stopped using Gemini. Hopefully it becomes better now.

Only available on Ultra, so I thought maybe I'll try it for one month to see if it's good. Well, I have Google One and for some reason this means I cannot sign up for AI Ultra? This is just one of the worst product rollouts I've ever seen

Website says its available on Pro too, but i have pro and can't access it.

I have the Free AI Pro plan from when I purchased my Pixel 11 Pro. I still don't have Argon in AGY nor web nor Gemini app. I have a free product and still feel cheated.

Likely because of phased roll out. Seems like a few others on the Pro plan are reporting the same thing. I am on the Pro plan and I see it in the Gemini app both on PC and Android.

I'm on the same plan (from buying a Pixel) and it's available to me

Feel better now?

Google's models are by far the best for readable text but also for nuanced understanding of the input.

Even after using the free model in Google search talking to Anthropic or OpenAI models feels like talking to an ox.

Gemini’s writing is really good, but unfortunately it hallucinates too much and doesn’t follow instructions very well. If only Gemini and Claude could be combined into one.

Ran on our Data Analytics benchmark: https://plotly.com/blog/gemini-4-argon-plotly-data-analytics...

Excellent model. Same price and score as Opus 5.5 in our 23 question exam.

This size / class of model is now a bit overpowered for these short-horizon data analytics tasks. OpenAI 6.1-Sol, Haiku 5.5, GLM 5.3 Flash all achieve about the same results but for ~1/10th the cost.

In other words, we're getting to the stage where models are "good enough" for certain tasks like data analytics and competitiveness is going towards size & speed.

Gemini 4 Argon

As of my knowledge cutoff, three of the top state-of-the-art large language models are Google's Gemini 2.5 Pro, OpenAI's o3 (and GPT-4.5), and Anthropic's Claude 3.7 Sonnet.

Knowledge cutoff isn't really what these models are used for.

Exactly my first test on all Gemini models. It cannot recollect any of the top events since Feb 2025. The worst about this is that since 3.6 Flash Google devs stated "knowledge cutoff March 2026" in the system prompt, so the model is wrongly confident that it knows about the year that it was not trained on. Which means that it searches less than it would.

And at least Flash 3.7/3.8 were both difficult to persuade that they have wrong knowledge cutoff.

I reported it as a bug to Google, but as all of my reporting to Google, no response ever. This is just stupid!

You need to use search grounding.

One thing to notice: Now, even in EU, if you turn on Google activity (so that you can see chat history), Google will train on your data. I remember it wasn't the case when in 2.5 pro period.

Put it through my standard benchmark, a six step explainer... yes, in 3D.

Very fast.

Makes mistakes. Corrects the mistakes quickly with direction.

Google Argon 4: https://share.gemini.google/T9r6tr2dDaTE

I didn't have time to make it into a full blog post yet. Here is a previous one.

https://generative-ai.review/2026/07/kimi-k3-rush-test-vs-cl...

For comparison here is the OpenAI Astra 6.

Astra 6 is significantly better than Google Argon 4 and a lot better than Claude Fable

https://generative-ai.review/2026/10/catchup-openai-astra-6-...

(edited thanks to tekacs)

I just created a Google account, which was a hassle. I tried to find an app and found AI Studio, but I’m not sure what that is. Then I found Antigravity, which apparently is Google’s alternative to Codex, and I installed it. Then I tried to buy the small Gemini AI Ultra package, but it didn’t let me pay. It accepted neither my credit card nor my PayPal account. It’s crazy how bad the Google experience is nowadays. You don’t know what to use or where to find it, and the security measures seem totally overkill. Am I really the only one having these experiences? Is Google aware of that?

I bought a google play card (or whatever it's called) with cash due to these issues. Result: Instantly rejected + locked after entering the code, no credit, no dispute, no customer service, no refund.

Yeah being presented with these options is such poor UX:-

- Try in Gemini

- Try in Google AI Studio

- Try in Google Antigravity

Will non users of Google products even know what these three things even are?

Same thing, I've always paid my bills, I have had some sort of google subscription for the last 10 years. Now they're requesting me to switch to pre-paid billing. They won't accept my card (that has worked every time this year so far).

Is there anybody in Google having a look at how bad and complicated their payment system is? In all the other AI providers you just sign in, pay, get a key, done. In other clouds like AWS, you add a card, pay little to verify the card the first time, monthly bill, done. In Google it looks like you have to fight them every time for the very thing they should make as easy as possible: opening your wallet.

I'll believe in AGI when Google can figure out how to leverage their model to make their user experience not suck.

Reminds me of that email from Bill Gates about trying to install Movie Maker.

I'm open to Google winning me over again, but there isn't enough hype in the world to do it. They will need to earn it after these disappointing 6 months.

I would love to try it out, and the Gemini models had previously been some of the ones I'd liked best in an overall sense, but unfortunately their updated inference endpoint is no longer ZDR on OpenRouter (change takes effect on the 18th). I think this is one of the more unfortunate turns in AI development in 2026, as it seemed like labs offering it was becoming the norm last year, and Google had previously done so consistently.

I'm a bit overwhelmed by Google's selection of AI environments (Antigravity, AI Studio, Gemini Platform etc.). If I just want the equivalent of OpenAI Codex, which one do I use?

As I approach a transitional period in my career, I'm reflecting on the managerial effort of places like google. How on earth do they keep an engineering org on the 'right track'? I work on products that are not as complex: fewer moving parts, fewer egos, no where near as novel. It feels like nobody does anything and nobody has any expertise

is it just 'Hire the right people'?

> The model launches today at an introductory price of $2 per million input tokens and $10 per million output tokens (with cached input tokens at 95% off).

Couldn't you just say what the price will be after the introductory period and when it'll take effect? Do they have some SKILL.md for these blog posts that requires the most important information to always be buried ten levels deep and hidden behind a login page?

Pretty insane that this isn't available for Pro subscribers on Antigravity, like not even a taste. Antigravity comes with a sample of Opus 5.5 too!

I gave the model my usual question of "I got a raise from X to Y, please explain what is the impact on my net salary". Answer was detailed and nailed all payroll edge cases, only to miserably fail in basic math at the end. It tried to calculate 430 minus 60 and got 410.

Announced two weeks ago and they went on a victory lap and they're already behind by the time they launched. Strange company, Google.

Adversarial programming language test for Argon. Surprisingly its safety guardrail gets triggered a lot just like for Fable 5.1.

    Claude Opus 5.5   58.7%
    Gemini 4 Argon    47.1%
    Claude Fable 5.1  46.7%

Benchmarks: https://bench.killswitch-lang.org/

Language info: https://killswitch-lang.org/

Excellent timing. I've been working on a UAV design and simulation tool and primarily using Flash 3.8 for the code gen grunt work because Claude hits weekly limits way too quickly. It's been working well this way. Flash 3.8 seems very efficient and i don't hit the limits before the refresh. Assuming Argon is similarly efficient then hopefully it means I can stop paying for Claud and switch over fully.

> Across Google and our early access partners, Argon is quickly becoming the preferred model for engineers tackling long-horizon software engineering.

"Across Google", wonder how much that is because of an effort by Google to curb usage of Anthropic models. MSFT and Meta similarly are pushing for in-house models. It doesn't mean the models are better when you push your engineers to dogfood and restrict access to other LLMs.

(Not that I know for sure this is the case, it's more of a hypothesis).

I have been very critical of Gemini in the past but one thing that I've felt Gemini has always been best at is being the most empathetic model imo. It feels far less cold and emotionless than GPT and Claude.

can't use in agy with Pro subscription:

Unavailable · Gemini 4 Argon is currently available to Ultra plan subscribers.

Excited to finally see this out. When new big models come out I like to make them audit any other code I generated to see how things are.

It's not available for people having a 20€ plan only. Am I just not a good customer with not. enough credits capacities ?

Right now there are two top posts one about GLM-Legal and one about Argon. Both score about the same in the Harvey Lab test, but the prices are $0.15 versus $10.55, the $0.15 is Vals AI's estimate using the base model since GLM-Legal has not published prices.

Maybe they could ask Gemini to fix their awful website to not crash on my phone!

Pricing is interesting. At launch Argon is the same price as GPT-6 Sol, half the price of Claude Opus 5.5 and 1/5th the price of Claude Fable 5.1 and GPT-6 Astra.

But on December 1st Argon doubles in price - so it will be the same price as Opus 5.5.

Gemini 3.1 is still SOTA at weird handwriting recognition. If Argon is better it will be amazing

All AI pages now include canvas animations. There isn't a single page without one. This is neither good nor bad, just an observation. I think OpenAI has the best animations so far, with Anthropic in second place. This one looks a bit off.

https://shot.3e.org/ss-20261013_160043.jpg google's roll-out procedures on point as usual

I'm a bit confused, all the announcements say that this is availabe to Ultra and Pro subscribers, but in antigravity, it's grayed out and says it's only available to Ultra.

Digging in the model card, "Gemini 4 is the next generation in the Gemini series, a suite of highly-capable, natively multimodal, reasoning models. Gemini 4 Argon delivers frontier performance..." If they're using noble gases there are 2 obvious lighter naming candidates and 3 larger (or 4 if you count Oganesson). I'm not seeing them claiming Argon is the top/biggest/bestest model among the Gemini 4 series, just that it has frontier performance.

Edit: noble gases not noble taxes, though the latter would be a funnier naming scheme.

I feel like if Google simplified their privacy/model training policy then their stock would go up.

It feels so complicated and hard to know what we can use vs what hands them our codebases

Gemini isn't perfect but I find it a lot less annoying than other model families (personal preference or stockholm syndrome). Interested to read the reviews.

Is it actually available?

Of all models I like to write with Gemini the most. The language appears much more natural and elegant than what even Opus 5.5 or Astra do. Also, they give 5TB of cloud storage and discounts, it’s a steal for what you get.

The code is bad though.

I do wish I could give it a try with some free credits but I would bet Google is wanting to limit usage until they know they can handle the traffic.

I'm starting to feel the subscription fatigue and a small amount of FOMO. I just can't justify pro accounts for every frontier model now that it includes OpenAI, Anthropic, Meta and Google.

argon seems to be really good at spatial reasoning, it's catching really minute details which other models like astra and opus seem to miss.

build this scene space station docking scene https://x.com/_satyam_ai/status/2110097633922662752?s=20

That web page has totally botched scrolling on iPhone. Quite annoying and distracting.

It shows a good improvement on my COCO object detection benchmark: 0.504 mAP zero-shot.

Gemini 3 Flash was around 0.403, 2.5 Pro was around 0.34.

not available. classic google.

I'm still waiting for a model which I can tell to "take these 3 screenshots and make a product launch video", where it doesn't randomly modify the screenshots or garble/rewrite the text as it moves the screenshots around the video.

Every time I try to generate one it's like a slot machine, where a different part will be messed up after fixing the first.

I hope all the people very concerned about AI safety are cheering this one. The best guard against one AI swarm taking over is having multiple adversarial orgs all with similarly advanced AI swarms.

I wonder if any of the AI safety people will begin encouraging AI researchers to swap jobs to work at a failing lab in order to keep them all extremely competitive with one another.

It's available for gemini pro users.

I think the model can be actually nice for co-working. I dont' know much about gemeni models but is this worthful?

Curious what this means for the anthropic ipo. Between this and meta, seems like there’s zero moat

It passed the test of using latest crystal-lang concurrent feature (with web search) when gemini 3.8 flash failed (again, with web search enabled). I think it is at least useful for coding, for sometime.

It still thinks Charlie Kirk is alive, though.

Here's how the 1M token output (long decode) works: https://ai.google.dev/gemini-api/docs/long-decode

It's nice to have it available for more complex tasks, but I'll stick with 3.8 Flash thanks to its high output speed. I really like it for fast iteration and prototyping.

not available yet. why to launch something that can never be used even after weeks

It seems the model is a little verbose, using more tokens to complete a task that other similar models as stated more precisely in the linked article about Artificial Analysis's work on this model.

Is there a benchmark for reverse engineering compiled programs? - Grader compiles a program to binary -> Agent has to reverse it back, and match all bugs and behavior.

So who has access to it? The say Ultra and rolling out to Pro, but are those the Google AI plans?

What about the Google One AI plans? Will they get it? That has a Pro AI plan as well.

I can’t try it because it’s not available on the free tier and no way in hell I’m tying a credit card to google, I’ll wait for argon-lite.

It doesn't matter if the model is good. Antigravity cli is unworkable, I thought auto permissions was a solved problem by now. I can't keep sitting here to press allow when in auto mode.

Here I was thinking Google had given up on frontier models...

The slides they show in the catering video demo are so bad (in addition to having all the usual signs of AI design), I would genuinely laugh if a supplier sent me this.

Finally, gemini 3.8 has been falling way behind opus and sol in my experience. It's showing up in my antigravity. My work account though still has 3.6, 3.8, 3.1.

at this rate we're gonna run out of stupid names for these things

"Hans, we're doing a new point release! What isn't crossed out on the periodic table yet?"

It's expensive. Google used to release good models that were cheap, but even the new flash models are expensive.

Wow, the site for the launch is horrible on mobile

Like it's a top tier model, so does that mean it is at the level of Opus 5.5 (claude) or 6.1 Sol (chatgpt)? I don't think I can trust the benchmarks Google released themselves, but from what users say, it seems like that. Since I already use both ChatGPT and Claude, I wonder if there is really any reason to also use this.

I am impressed by how fast it is. Haven't run all my evals but it's surprisngly quick for creating files

> Our next era of frontier intelligence, built to solve your hardest problems

On safari the page doesn't scroll very well on the first section

Any word on how fast it drains usage compared to openai/anthropic sub for the different tiers? I guess too early to tell?

Google has too many options that may confuse users. I don't know if I should use Antigravity or AI studio or Argon for development

Which one is it?

So, Apple? If Siri uses Gemini, when is this Gemini 4 coming to Siri and how much is it going to cost us?

The previous gen pro used to have this weird tick where if you ask it a reasonably open ended question, it would number things in the response and assign each one some weird label starting with 'The...'

e.g.

    X: The Parameters
    Y: The "System"
    Z: The "Ongoing Grind"

The first two prompts I tested in AI Studio does the same, unfortunately.

Another quite unpleasant tick was that if you specify a peculiar term in your prompt as a conversational shortcut, it would keep repeating that throughout the conversation. e.g. "...What is a bitter lesson pilled way to do this?", then you'd see it will quote "bitter lesson pilled" at like 7 places in its response. This one seems to have eased off a little.

Currently unavailable to ultra-plan subscribers (on agy)

Impressive! Previous to Argon I instructed Claude to find preimages in 19 rounds of SHA-256 (with padding) [1], then tried to move to 20 rounds (no padding) where it couldn't beat ~2^45 table lookups, and Argon found preimages in ~2^33 (no padding) [2].

[1] https://github.com/srw/sha256-19step-preimages-with-padding

[2] https://github.com/srw/sha256-20step-preimages-with-no-paddi...

The hardest problems, but not yet a website that can scroll on an iPhone 17 Pro without stuttering reflow.

Does Google have a subsidized plan like others? If yes, is it only eligible for use through Antigravity?

Doesn't seem like anything special in my tests, not very different from the previous Flash models.

My first experience in antigravity was extremely disappointing.

I asked for a QUICK healthcheck on the sync status of a bigquery integration. It literally spun for 25m before I killed it.

The task, which is CLEARLY outlined in the agent readmes just involved using the gcloud cli to initiate auth, then just run a few queries to check logs and timestamps.

Even 3.6 flash can get this right (most the time).

I made a NHL inspired hockey web game with Opus 5.5, and I let Argon loose on it to try and see if it could improve it, and it found a ton of bugs and improved performance and load times significantly. If anyone wants to play it here:

https://playblueline.pages.dev/

Not available in my browser/app yet (pro), but I'm confident it'll roll out soon.

That was a long wait.

I tried it out in the chatbot and it is significantly better than previous models; when will it be possible to use it with Antigravity if I have the PRO plan?

Can’t even get it on pro tier in antigravity?

What’s the point lol. I’m not subbing to ultra to have it

The should ask Claude to make the page scroll faster.

Does this mean Googlers will stop using Claude internally?

Do we need to post every single model release on HN?

Does this solve the 2 year old bug of recitation hell?

Gemini 4 argon is horrible on mobile because of the header it takes the half screen

Tried it today on some relatively complex mid-sized real world tasks with several thousand lines of code generated by Argon. My impression is not good. Maybe it's due to first-day high user counts, but it seems to be doing worse than 3.8 flash in some coding tasks. What I noticed was that it often forgot significant parts of detailed, well-structured and precise instructions (which is somewhat rare with newer Claude and ChatGPT models) and not just for small details. It also seems to spend too much time thinking even when the task was a quick and simple one.

Without going into details, for a small part of a larger task, I asked it to classify different content types of different files and formulate a structured mapping (e.g., in JSON) between content types and another set of files (deliberately left vague here but not to Argon) from my codebase. I even offered it a multi-part draft of a classification system to start with and told it that if it can find other classification systems from international standards, academia or industry more suitable to the task than my rough draft, we can discuss replacing my suggestion. There were other parts to the instructions too, written out like an article with bulleted lists, precise program-like wording, etc. Then, I sent the prompt and left for a while.

When I came back, it was nearly done with the implementation but the classification system it came up with was a mess. It even admitted to making up "bogus" classification standards when I pointed out some errors. It also kept leaving gaps multiple times when I referenced errors to fix in subsequent instructions (and not once apologized the entire time lol). So my first impression was subpar, not really comparable IMO with Opus >=5 or Sol >=6, and seemed to leave larger gaps than 3.8 flash typically does. The most irritating part to me was not only the fact that it was making up things, but that it ignored my instructions to discuss things with me before deciding to shift away from my draft; I don't recall any major LLM ignoring an instruction like that in at least a year.

After having de-Googled and reading people’s experiences trying to run Argon 4, I think I’ll pass.

Now if only I could opt out of training without disabling chat history on a personal plan like I can with ChatGPT and Claude...

Looks at watch.

OpenAI release any time now

Not available for me on Pro in the UK.

Just switched, and it seems to be much better. I'm hoping it stops lying. I have Pro max subs for Claude, Codex, Grok, and Google AI... I was forced into using Google by running out of credits elsewhere.

I have found that Gemini often claims it did something, but didn't actually complete the task.

I just had 3.8 Flash do a major refactor -- that never should have needed to be done in the first place -- and switched to Argon 4, and had it review it.

Almost everything had been stubbed out by 3.8, and none of it was properly integrated into the codebase. :facepalm:

Unlike most people it seems, I like all the progress in AI but I really do wish they would reign in the over-the-top marketing. I mean, can we just get this stuff to the point where I can reliably work with it to design a well-architected software project without having to constantly struggle against bloat and nonsense directions no experienced or sane developer would ever take.

fyi: argon is an interesting choice of name, coming from the Greek argos meaning "lazy or living w/out labor".

So like... do I get access?

And yet, for all of this, I cannot scroll this website without stutter and lag on my (current gen) phone!

Is it faster? Thats the only thing that matters for me anymore. Models are probably as good as they ever need to be for coding, but please make them faster, please.

Now when another uninteresting frotier model been released Google can finally start doing what we actually waiting for:

Start preparing Gemma 5 release.

Glad to see them abandon the absolute train wreck of their model naming "strategy".

Gemini Live vs. Gemini Lite vs. Gemini Pro vs. Gemini Flash vs. Nano Banana Lite (that one apparently has no Gemini prefix?)... Please, just pick a metaphor and stick with it without wildly mixing reasoning modes, model sizes, and input/output modalities in the model name!

isn't harness as important as model these days?

I wasted $20 in credits to test it, thinking it’s better than Codex Astra or Sol. Nope. Disastrous by comparison.

Only available in the $100 plan, and it's worse than Opus 5.5? Why would you ever buy this?

Mindboggling output window. I'm kinda jealous, I used to use gemini til google banned my 20+ year email for "bot-like behavior".

Still reading up on it here: https://ai.google.dev/gemini-api/docs/long-decode and needs lots of community testing to determine whether it's a "true" 1M token output window.

Fable 5.5 about to drop the hammer on them and then another Astra version goes out lol.

The free version now uses Gemini 3.5 Flash-Lite rather than Gemini 3.6 Flash.

Going to take a lot for me to move from subsidized Claude opus 5.5 … this Google model is very cheap though.

Is this supposed to be the new Gemini Pro?

can for the love of god we get some usage in cheaper subscriptions being Ultra tier only is annoying. I want to just test it out here, well will burn some money via api I guess.

I've asked Argon to do 3D model for me. It said that it can't do it. I've linked promotional materials and said that it actually can. It agreed and got me a model (not a great one). Man, why is Gemini like this? Every time I fall for the hype I get disappointed.

I use Gemini all day at work via the chat interface. It's rock solid. I find it hard to understand which model it's using though. Am I already using Argon, or does it switch models according to the subject?

not available in our org (EU)

>"Argon helped the Chrome team separate core webpage handling from a ~6 million line legacy folder of browser features held within Chromium’s 40 million-line codebase. Piping multiple clean code changes at a time, Argon helped the team finish an architectural refactor across hundreds of files, completing months of work in weeks."

Now that needed to be done for some time!

Regardless of if an AI helped or not, regardless of whether that was Google's or some other AI or not (though Argon does sound pretty cool!), regardless of whether an AI cut the time from months to weeks or not, refactoring Chrome -- decomposing it into individual clearly demarcated non-tightly coupled software parts with core webpage handling at the center -- that needed to be done for some time!

Now all of that being said, Gemini 4 Argon does sound pretty cool, and we congratulate both teams at Google -- the team that created Gemini 4 Argon, and the team that refactored Chrome with it!

But all of that being said, while the advancements made for Gemini 4 Argon are important and impressive (and they are!)... refactoring Chrome properly (IMHO) is slightly more important and impressive!

(You know, one man's humble opinion! That is, I am not the guy who is responsible for awarding performance-based bonuses at Google at the end of the year, so take everything I say with the proverbial "grain of salt"! :-) I am probably biased, one way or the other! :-) )

Why can't the Gemini model code? I'd guess Gemini probably used the best dataset. Even Google is the place where the world's best coders are gathered, isn't it? Yet the GPT code from OpenAI, which collaborated with Microsoft, is beautiful, while the Gemini model's coding is trash.

Are they gonna go to Krypton or Neon next? Kinda started in the middle there.

Meanwhile I paid a subscription and they only provide access to 3.6 Flash. Consequently, I have moved on to open-weight models. It is what happens when you screw your customers over.

3.1 Gemini was amazing for anything non-agentic. I switched to it for my primary chatbot purposes.

I never touch 3.8, it's awful. Curious to try this, but I've yet to use any Gemini that is good for agentic work.

Also, unrelated but this made me look at Artificial Analysis to see where it lands on Cost per Task, and it ranks Opus 5.5 at almost double Astra 6. That does not align with my experience.

People use Antigravity still?

Not available for AI Pro subscribers in Antigravity. What a disappointment.

Google...why. You keep getting so close but constantly fumble at the last moment. I can use Opus 5.5 and Sol/Astra on the $20 Anthropic and OpenAI plans.

You are already playing from behind. You need to GET YOUR MODELS INTO MORE PEOPLES HANDS by any means possible. Continuing to make it difficult to use your products DOES NOT HELP.

This model might be great but the majority aren't going to subscribe for $100/month just to try it.

very bad

baaad

why they wait so long ?????

Now let me use Pi on the plan.

It may be a good model, but I'm not reading scrollslop. It annoys me to much when I'm just trying to get information about something.

so this is prob why antigravity is not working well at the moment lol

Great. More AI slop

That's why Gemini 3.1 Pro has been shit for weeks now.

Blatantly obvious that models will not slow down. For any task, if we can build a dataset or environment of verifications, it can be automated. This will eat all of human tasks.

The fact that multiple companies can breach any benchmark means there is no ambiguity in terms of how to build AI, its just a matter of time.

Certainly Meta and Google are benchmaxxing to please public markets, but it still proves the above

> built to solve your hardest problems

Like a working scroll on a webpage.

God damn this is impossible to use of safari mobile.

Good God, man. Google has the worst frontier lab rollout of them all. And they're delivering products that are behind the curve constantly.

Now, I happen to believe that's an excellent strategy. I bet they actually resolve more LLM requests than any other company at this point. The lay person still uses Google exclusively.

To everyone saying not available yet: it looks from the page like Pro and Ultra subscribers have it today and given the play links, it'll probably show up in the API in a few hours.

Classic Google: Clicks the "Try in Gemini" button from the Gemini 4 page, gets a page that defaults to Gemini 3.6, with no option to try Gemini 4 without upgrading.

If they don't want to make it available even for small tests for free, that's fine, but then don't put a button suggesting I can try it without qualifying it...

EDIT: It's actually worse: The page then shows "Look what you can do with Gemini 4" and a "Let's go" button that leads to a bunch of links to examples, all with enticing pictures. Except if I click on them I get a box stat strongly suggests it'll run with Gemini 4, except the dropdown shows "Flash" and if you click on it, it shows as Gemini 3.6 again, and if you run it, the results unsurprisingly end up looking nothing like their pictures.

I don't know if that page was built by humans or AI, but clearly nobody tested it properly.

A model that didn‘t break out of its sandbox and hacked other websites cannot be good.

I'm pro and I see that I need to be Ultra to use Argon :/

Meanwhile, here's Gemini in practice:

Me: Gemini, navigate to nearest McDonalds.

Gemini: [Lists 10 McDonalds locations wnd asks which I want.]

Me: What were the priginal instructions just now?

Gemini: [Recites imperatives about tone and content from its memory system.]

Me: No. I told you to navigate to the NEAREST McDonalds. There's only one NEAREST McDonalds. Do that one.

Gemini: My mistake, navigating to the nearest McDonalds. [Opens Maps, navigating to a McDonalds across town.]

How is Claude Opus 5.5 beating Gemini 4 Argon on benchmarks but everyone here is calling Argon a step change? I need to come up with my own benchmarks ...

The Google Gemini subscription landscape is such a mess... gemini.google.com welcomes me with "Look what you can do with Gemini 4", but the model selection is 3.6 Flash, 3.8 Thinking, and 3.1 Pro...

Reply on news.ycombinator.com