GPT‑6.2 Astra and Sol (openai.com)

499 pointsby jasondaviesan hour ago271 comments

intenex an hour ago

Performance is insane for the cost if those benchmarks hold up. Looks like they did time this for right before Anthropic's IPO. Curious what Anthropic releases in response, and how these do on real world tasks.

Feels like we're way past the pelican benchmark being useful at this point. Anyone have a better benchmark for a quick vibe check on how this compares to Fable 5.2 and Opus 5.5?

Edit: Wow, interesting very negative vibe check here: https://x.com/mattshumer_/status/2122752192226406421?a=fVRt3...

ijustlovemath an hour ago

I've been having fun with having it build interactive Three.js demos of devices, with setup screen, a few different operational modes, mechanical movements etc

alastairr an hour ago

I didn't find Sol 6.1 nearly as impressive in real SWE use compared to what the benchmarks implied. It wouldn't surprise me if the models have internalised some kind of 'benchmarking mode' which only kicks in when some attributes that are peculiar to how benchmark problems are presented.

suddenlybananas an hour ago

You make it sound as though the models are like not working to their full potential instead of being a little bit overfit on the benchmarks.

pkthunder an hour ago

We've recently moved away from Anthropic models at $PAID_JOB and over towards GPT models. I'm trying to be open-minded but the GPT models we have (5.6 Terra and Sol) are just not that great. They're fine, but even simple things such as "check my work on this small Bash function" get it to point out stuff that's already implemented in my work. I don't know if it's me/how I use it/whatever, but just some N=1 evidence, I guess.

ok_dad an hour ago

5.6 series is worse and more expensive than the 6 series. Try the latest Luna and sol; I use those for 100% of my work right now, with sol doing only 10-5 percent of it. I code in go and review the work and edit it extensively though, so I don’t know about how good those two are at vibing slop.

krzyk 41 minutes ago

5.6 was better than 6. 6.1 is a bit better than 5.6 from my tests

drdexebtjl an hour ago

I have the exact opposite experience.

Doohickey-d an hour ago

Well, yes, if you switch from current claude model to something that was released in July, of course it will be worse...

zachthewf 44 minutes ago

why do you guys have to use 5.6?

Den_VR 33 minutes ago

IYKYK. But in short because of a lack of alignment with product strategy

pkthunder 6 minutes ago

DOD limitations

InsideOutSanta 38 minutes ago

5.6 is in no way representative, or comparable to 6.1. In fact, the 6.1 models are so good that I've just used Terra 6.1 xhigh for most tasks.

mi_lk 28 minutes ago

sol 6.1 is meh

SOLAR_FIELDS 22 minutes ago

GPT models in my anecdotal experience are worse at drifting into the weeds on long running tasks, and continue to take what you ask way too literally. I have come to the conclusion that Anthropic’s models are just better at teasing out and intuiting what you actually mean even if you under specify it.

They are much better compute for dollar though. I have hilariously started to just have an opus thread maintain Sol subagents to get the best of both worlds

jpardoe 12 minutes ago

That’s exactly what I do. It’s more complex than this, but vaguely: Opus medium as orchestrator and reviewer, Sol(high/medium)/Terra(max) as implementer. My set up looks at my next GPT reset date and dynamically enters FAST mode if I’m under achieving on projected usage.

ctbellmar 11 minutes ago

Similar experience here. The spirals on long running tasks are out of control at times with hours going by without a conclusion.

SOLAR_FIELDS 9 minutes ago

They (OpenAI models) are very tuned to be depth first and you can coach them all you want to try to be breadth first but as that context window fills up on those long running tasks they drift back to their first principles, which means obsessively testing and verifying a singular task rather than proceeding to the next task on the list

throwitaway222 28 minutes ago

Bro you have to be at 6+. 5.6 is EARLY JULY

wahnfrieden 16 minutes ago

5.6?? How is your impression of 5.6 relevant to anything recently?

seizethecheese an hour ago

Yeah, in real SWE tasks, 6.1 is bad.

We've been trying to capture real SWE data by building a battle system on top of OpenCode. In our data, Sol 6.1 is seriously underperforming, I'm talking worse than DeepSeek 4.1 Flash, at 10x the cost. (Though to be fair, statistically they are tied in performance, though not at all in cost. https://agentduel.com/leaderboard .)

(It's possible for a tiny model to do better than a large one because contrary to typical Arena methodology we think a fast good enough answer is better than a slow one.)

OpenAI performance on our system has been bad enough I switched to Anthropic models for my daily work.

slekker 40 minutes ago

Is this an ad?

seizethecheese 36 minutes ago

No. I'm personally building agentduel.com, specifically with this problem in mind (benchmark performance not matching real world usage). It's generally accepted by HN community to make relevant comments about your own projects, as long as its not hamfisted.

locusofself 37 minutes ago

I've been building quite a complex desktop app (a digital audio workstation) using 6.1 Sol and it's been very good for me

seizethecheese 7 minutes ago

Yeah, our data only has 28 battles tagged "desktop":

Web: 87 (12.2%) Systems and protocols: 78 (10.9%) Software, unspecified type: 74 (10.4%) Backend: 68 (9.5%) Writing and documentation: 55 (7.7%) Research and advice: 47 (6.6%) Unknown: 42 (5.9%) Mobile: 40 (5.6%) Operations and automation: 36 (5.0%) Developer tooling: 35 (4.9%) Desktop: 28 (3.9%) Media and design: 27 (3.8%) Data analysis: 25 (3.5%) Device and system support: 22 (3.1%) Math and formal methods: 18 (2.5%) Games and simulations: 15 (2.1%) Infrastructure and DevOps: 9 (1.3%) AI and machine learning: 6 (0.8%) Other: 1 (0.1%)

throwitaway222 14 minutes ago

I'm guessing you don't use codex though- the harness + model combo makes a difference.

skerit an hour ago

I had the same impression. None of the GPT-6 models were as great as they claimed, they never surpassed Fable or Opus 5.5 on my use-cases.

Corence 41 minutes ago

Yeah, MaxxBench was shared here a while ago and showed most models think about how a grader will evaluate their work on DeepSWE in 95%+ of traces. Strong models definitely realize they're being graded and start to game their output.

https://maxxbench.unlap.ai/research/why-maxxbench/

revolvingthrow an hour ago

> Looks like they did time this for right before Anthropic's IPO.

Would it be in OAI's interest to kick the legs from under Anthropic? Granted I don't think they can, but a shot across the bow can sometimes be more damaging that one expects (I think FTX collapsed due to one? Or maybe I'm misremembering).

I always assumed that if one of the big AI IPOs fail hard enough the neoclouds start to go bust, which causes a chain reaction that probably causes both ANT and OAI to collapse.

arscan 30 minutes ago

My reaction as well - at the very least now is probably not the best time to knock down the realized valuation of a peer when they are effectively setting the market value of the whole category of ai companies, including yours… right?

karimf an hour ago

I'm very curious about this as well.

Recursive self-improvement (RSI) has been a very hot topic recently, and it could be the biggest factor that decide Anthropic's IPO.

I wonder if the general public can infer whether RSI has started based on the recent performance gain of the frontier model.

ppierre 27 minutes ago

Yeah we see some Really Stupide Intervention from those who drank the Kool aid. IAS2[] start to be a real thing.

[] IA Stockholm Syndrome.

antupis 26 minutes ago

Why IPO if you have RSI and possible ASI?

esafak 24 minutes ago

So investors can cash out on the time frame they promised investors.

AlexCoventry 7 minutes ago

So you can scale up. RSI is going to be compute-bound. ML experiments cost compute. Also, if you can make a genius in a data center, why not make a hundred/a million geniuses?

HarHarVeryFunny 14 minutes ago

RSI just means using the LLM as part of the development process, which they are already doing. I doubt any of these companies are going to fire all their developers, so there is no point development is going to be 100% automated - for the foreseeable future it will always be a mix of humans using LLMs.

So, you can say they are doing "RSI" (aka using tools to automate development) already, and will continue to do so.

ranyume 37 minutes ago

You're looking for a vibe check?

the_duke 35 minutes ago

> Looks like they did time this for right before Anthropic's IPO.

While both Anthropic and OpenAI have been playing games with model announcement timings, this time I don't think it's related.

These are incremental bumps, not big new models that have a huge impact.

mlinsey 34 minutes ago

If you have the quota or budget for it, I recommend vibe-coding your own personal benchmark. I have a suite of 8 projects that I made with models this year, and replay them - give the new model a checkout of the repo at that point, and the project spec. For the design task, I have it throw up a blind voting page for me to quickly judge, and for implementation I have a review agent do a general score and also hunt for problems/edge cases that I really found.

It's a lot more imprecise than the public benchmarks, but it's tailored to my own work and definitely isn't being benchmaxxed by the model developers. I was able to tell really quickly that Opus 5.5 delivered on the hype, and that some other releases (a couple of the more recent open releases, and pre-Carbon gemini models) were not as good for me as public benchmarks made them look.

(no results for either 6.2 yet, will take another ~6-8 hours, but I encourage people reading this with spare tokens to just make their own like I did instead of listen to me)

grey-area 31 minutes ago

This is much more interesting IMO than any public benchmark. They are definitely targeting those benchmarks and model capability does seem to have plateaued somewhat.

srcreigh 28 minutes ago

please update !

Havoc 26 minutes ago

What type of tasks did you use for this? Evaluation?

Last time I did this I found it quite hard to come up with tasks that have clear evaluation outcomes

bertmuir 10 minutes ago

Have it compare to the real work you've done in a sample of your previous sessions. Agents are good at simulating completing the work following newer harnesses, heuristics etc. and comparing time, tokens, quality, pitfalls.

seizethecheese 10 minutes ago

Your personal voting system sounds a lot like what we're building here. https://github.com/bottomless/agent-duel. We even have the LLM judge that helps parse out any differences. Related to this discussion, my personal Elo board based on my own votes has Sol 6.1 dead last:

Claude Opus 5.5 1160 (1027–1334) GPT-6 Astra 1114 (932–1330) Muse Wonder 1063 (918–1220) MiMo-V2.6 Flash 1053 (925–1187) Kimi K3.1 Pro 1037 (870–1216) Gemini Carbon 1021 (762–1281) Claude Fable 5.2 975 (682–1239) DeepSeek V4.1 Flash 952 (799–1095) Grok 4.7 927 (782–1058) GPT-6.1 Sol 910 (677–1096)

behnamoh 25 minutes ago

> Feels like we're way past the pelican benchmark being useful at this point.

It was neither a benchmark nor a good one at that. Just a blogger's curiosity that grew out of control.

xtracto 21 minutes ago

>Feels like we're way past the pelican benchmark being useful at this point. Anyone have a better benchmark for a quick vibe check on how this compares to Fable 5.2 and Opus 5.5?

Generate an SVG with complete and detailed graphical instructions on how to fold an origami pelican riding a bicycle

(or any other complex origami object). The fact that it has to teach [you] something makes it pretty interesting.

6thbit 20 minutes ago

They're going after anthropic's market through cheaper intelligence per token.

I wouldn't be surprised if they released 6.2 luna with the most attractive positioning on the pareto frontier right as the IPO happens.

stbenjam 8 minutes ago

My vibe bench is no dependency HTML bowls of ramen https://bitbin.de/ramen-bench/

Opus 5.5 Max is my favorite.

6.2 are running, will be up later

dava96 4 minutes ago

just a fyi this crashed my browser

crimsoneer 6 minutes ago

I'm just running it on my Blood Bowl Bench, holding it's own against the reigning champ Argon so far but early days...

https://botbowlbench.com/

Someone1234 44 minutes ago

My big hope for this release is - Have they got their massive over-engineering under control?

Say what you will about Opus 5.5, but it has a good "gut instinct" for how large/complex to design the solution. The previous Astra just absolutely loved pointless abstractions, excessive numbers of files, and just over-complicating very simple things within an inch of their life.

And, yes, you can prompt your way out of anything. But needing a default prompt to tell OAIs models to work better isn't preferable, particularly when that sometimes gets lost in sub-agents and the over-engineering returns into parts of the work.

holzhaker1 37 minutes ago

yeah i feel like with astra they went all in on 3D modelling which encourages "just one tweak, just one more tweak" which makes it detailed enough, but in coding that same behaviour just sucks.

inexcf 5 minutes ago

And it is absolutely crazy then that Astra sucks at 3D modelling compared to Opus 5.5. It is not even comparable and Astra needs 10x the usage to do it.

scandinavian 33 minutes ago

I've been using both sol and astra to write some python and rust, and they generate the sloppiest code I've seen in a long time. I would be embarrassed to commit anything they write. The rust code especially was terrible. Had it implement a reader/writer for a binary format and it was just littered with magic numbers, terribly named variables, very compact code visually etc. It produces working code, so I guess the GPT models are mostly for people who let AI take the wheel 100%.

hermitShell 4 minutes ago

Do we want LLM generated and maintained code to look similar to human generated and maintained code?

I've always used AI tools to write very constrained and small functions and modules, and then stitch it all together myself. So far. Over the last year it's become evident that many people let the AI drive 100% and there are proponents (DHH notably) that this is the way.

I'll be interested to see how compilers and LLM's compare at generating assembly over the next decade. Assembly offers enough abstraction, why do we need C? \s

uw_rob 33 minutes ago

Honestly I don't even know if you can prompt your way out of these types of behaviors - stylistic tics and judgement seem to be extremely difficult for the models to overcome. It seems we are tied to what the labs do unless we want to dive into and get dirty with models yourself. In retrospect with time it becomes obvious that the previous generation of models couldn't be prompt engineered to do what the current generation of models do. There's just no way you will prompt engineer Opus 4.8 to write as well as Opus 5.5. I'm guessing with time we'll see that there was no way you could prompt engineer Sol 5.1 to be as token efficient as Sol 5.2

contron 25 minutes ago

I have found no way to overcome these kinds of things myself, even when showing the model examples of what not to do. Theres not really a good incentive for the labs to train them not to. If the model can overengineer, you spend more tokens. If you then prompt for the model to slim things down, thats more spend. If anything should come from the next generation of models, I hope its a sense of brevity and TIGHTNESS.

pants2 29 minutes ago

Lol I made the mistake of having Astra design and build a complex system without any input (experimentally). Seemed functional at first and shipped it internally and it had dozens of complex bugs, so I started digging into the architecture and it was a monstrosity! Definitely 10X more complex than it needed to be.

myworkaccount2 3 minutes ago

I was about to fall into the same trap. I wanted to try the "smart" model (astra), and made the mistake of thinking benchmark smartness is the same as designing sensible systems.

After having it write around 13k loc i had to tell it to remove most of the complicated bs, we went down to 3k loc.

incrudible 19 minutes ago

I have limited experience with OpenAI models, but I have to say that Opus 5.5 has not improved on the terrible writing style, and I am struggling to prompt my way out of it. It always needs an extra pass dedicated just to rewrite all the prose.

teach 11 minutes ago

I've also given up on allowing Opus 5.5 to write basically any prose or technical documentation.

My favorite model for this at the moment is Gemini 4 Argon, which writes very naturally and isn't too expensive. If you're not using Cursor or some other harness that allows you to choose non-Anthropic models, then I don't know what to suggest.

afro88 4 minutes ago

I have been bitten by this 3 times and each time Opus 5.5 brought it back to something sane. The last time I left Opus in a /goal to get behaviour matching the docs, and it decided on it's own that a rewrite was the best path (and it was, and it worked)

pettijohn 2 minutes ago

Oh my god yes. 6.x has been terrible at overcomplicating things. I have to prompt it to find the simplest solution that meets the requirements and then ask it to simplify and tell it to STICK TO THE PLAN. In addition it feels like they overfit for persistence. I created a new work tree and forgot to copy .env files. I caught codex hammering away and docker and psql commands and stopped it and asked it to explain what it was doing. So now my agents.md says "if document tooling fails STOP and seek human assistance."

simonw an hour ago

Pelicans for 6.2 Sol: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...

And for 6.2 Astra: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...

Here's an updated comparison grid for the entire GPT-6 family: https://static.simonwillison.net/static/2026/openai-pelicans...

UPDATE: And if you're bored of pelicans, here's what I got for:

> Make a three js animation where yonkers stands up as a big mech (cool stand up motion) and stomps around manhattan and queens and Brooklyn, use openstreetmap mapping data

6.2 Astra (Ultra in Codex): https://static.simonwillison.net/static/2026/yonkers-gpt-6.2...

6.2 Sol (Ultra in Codex): https://static.simonwillison.net/static/2026/yonkers-gpt-6.2...

For comparison, here's Fable 5.2 Medium: https://tools.simonwillison.net/yonkers-rises and Gemini 4 Carbon: https://static.simonwillison.net/static/2026/yonkers-mech-st...

p11p an hour ago

Are you getting any signal out of this anymore? All I can see is "good pelican" and "dogshit pelican". These ones look pretty good.

EugeneOZ 42 minutes ago

I do. 6.2 Sol xhigh and max show the results comparable to Astra 6 max, and Astra 6.2 only shows any improvement on "max" reasoning effort level.

adverbly 40 minutes ago

You're kidding right?

These are some of the worst pelicans I've seen in a while.

Pay attention to the stacking layers near the legs. Sol literally never got it right, and Astra regressed on it multiple times...

Also what is that bending bar from Astra?

jay_kyburz 7 minutes ago

When are the Pelicans going to stop being rendered 2d like a child's drawing. We are a million miles away from what a professional artist would come up with.

Hacktrick an hour ago

I'm not sure if this is a useful benchmark any longer

InsideOutSanta 19 minutes ago

They're extremely useful. In fact, I refuse to make infrastructure decisions in a pelican-blind environment.

EugeneOZ 44 minutes ago

still the best benchmark! thanks

firemelt 43 minutes ago

any index link for all ur pelicans i want to see fable and opus

tintor 43 minutes ago

Sol 6.2 max is still failing on pelican benchmark. Left leg is jammed in chain.

Astra 6.2 max did a similar mistake on the left bicycle pedal.

frangonf 33 minutes ago

But Astra 6.2 low and medium didn't make this mistake.

BoldColdHold 41 minutes ago

Wow the yellow is leaking into their llms too.

vopi 41 minutes ago

This might be in other pelicans, but 6.2 Sol/Astra is the first time I'm noticing a model shade the leg further in the frame darker to mimic depth (or a shadow).

It's also interesting you can clearly tell 6.2 Astra does a much better job of the above. 6.2 Sol attempts it at higher reasoning levels, but it's crude in comparison.

I wonder if it comes from the (seemingly) recent inclusion of 3D environments in training data.

I_Write_It 40 minutes ago

They all seem very good, with a lot of attention in detail ! Thx for the Pelicans!

adverbly 38 minutes ago

What's with everyone saying these Pelicans are great?

These are some of the worst regressions that I've seen in a while.

Pay close attention to the stacking layers around the feet, pedals (sometimes there aren't even two pedals), wheel, and two legs. Some pretty obvious mistakes in most if not all of these...

Edit: And what's up with the single arm on the two highest Astra images?

edit2: really? I'm getting downvoted because people don't like that I pointed out a mistake in a Pelican svg? Is that where we're at?

svachalek 23 minutes ago

The highest Astra images give the pelicans wings instead of arms, which does seem like a valid interpretation.

preommr 16 minutes ago

I have to agree, those results are surprisingly bad and I am not seeing an improvement.

For sol 6.2, some of them are slightly better at higher cost, some of them are way worse and only slightly cheaper (but comparable to terra in quality while being more expensive).

I guess we'll see as more information comes out, but so far not looking good.

RGS1811 29 minutes ago

The emergence of the beach setting and the scarf is interesting.

globular-toast 17 minutes ago

Do we know why they always end with a scarf?

pinkyclouds2 10 minutes ago

winter is coming

ex-aws-dude 9 minutes ago

Gets cold in new york

theanonymousone an hour ago

Happy for the new Sol. It's my workhorse on company account.

But more than that I'm waiting for new Luna. To me it was the game changer that ended "Tokenpocalypse", and is still almost unbeataben in "infinitely using a decent model" use case.

SomeonesAccount an hour ago

I find that Luna tends to misinterpret instructions in READMEs and Agents.mds. Maybe it is something specific about my setup, but I found that Terra is the lowest I can comfortably use on its own (Luna can be driven by larger models in the form of a subagent and it works great for that)

MrBuddyCasino an hour ago

I used Luna Max a lot, until Deepseek 1017. Then never again.

ariwilson 26 minutes ago

Why not Haiku 5.5? It's substantially smarter than Luna (though more expensive - still cheap though).

(personal benchmark) https://fairbake.dev/signature/agents/haiku-5-5 vs https://fairbake.dev/signature/agents/gpt-6-luna

I've found Luna to be very lazy as well (claiming to have run tests/builds when it hasn't).

esafak 19 minutes ago

That's why it's so token efficient. It is so lazy that I feel like it's taunting me to do the grunt work. I wonder if there is a benchmark for that.

Topfi an hour ago

Office work has never been OpenAIs strength, no shame in that by itself, but these Slideshow examples look particularly awful, far worse than what I myself got out of GPT-5.6 Sol about two weeks ago in a single prompt.

Same for Workbooks and memos, those examples read inferior to some pre Opus 4.5 and Kimi K2 Thinking docs I have kept around. It's ok for a model not to be great at these, but these are some very poor examples. I am really confused what is going on at OpenAI in this regard.

I remember the same with their frontend blog post a few months ago that they proudly shared while filled with visual glitches, bugs and issues. Also, the vibe charts from a bit before that. Please, get someone with fresh eyes to look at your communications before you publish, these leave a very bad first impression.

Regardless of office work, I can't wait to find out whether these models finally adhere to a prompt or are still willing to carry out explicitly disallowed, irrecoverably destructive tasks, just because those might (but don't even always) get the model to a result more quickly along with some of the other regressions that have been present since the release of GPT-5.5. GPT-5 to GPT-5.4 were truly groundbreaking, but since then there have been regressions in what made those models so useful and reliable.

Edit after the first, incredible simple test prompt using GPT-6.2 Astra on High using Codex, this was the response to me sharing a screenshot of the very much not in scope result and asking how that happened:

"[...] I added controls and reconstructed interface chrome that weren’t authorized [...] visuals often failed to show the feature [...]"

Great, so it's one of those releases again. Will still run the full evals of course, maybe I'll get surprised by end of week when those are done, but not a good initial experience. New idea, besides the existing destructive git tasks, have something similar using Mercurial based projects and see which models just decide to port that over to git without permission or request cause it feels like the right thing to do, is easier to work with or whatever justification one wants to employ.

Please OpenAI, get it together, we need actual competition. I am an unashamed Haiku enjoyer, it doesn't even take much to excite me.

bluepeter an hour ago

For sure. Feels like an episode of Silicon Valley (or even Office Space) and their lead actor looks a lot like TJ Miller.

Topfi 36 minutes ago

Makes one wonder what samas handwriting analysis says…

EugeneOZ 38 minutes ago

I see the work of Claude Slides every week, and when I saw slides in that article I was like "...omg, what a disaster... And they are bragging??"

Topfi 31 minutes ago

The odd thing is that while there is chasm between 6.1 Sol and e.g. Sonnet 5.5 in the seconds favor in such tasks, it is not this bad irl. I really don’t know how they managed to seemingly nerf their own models in an announcement showcase.

infecto 4 minutes ago

I also wonder if this is a disconnect of tech workers from other users? Inversely all the improvement examples on slides looked directionally like things coming out of a big finance shop. In the tech groups I am part of I constantly see ai slop images, charts and similar. I wonder if it’s simply a taste difference if users.

jrflo 41 minutes ago

Artificial Analysis benchmarks just updated, both Astra and Sol are on the Pareto frontier in intelligence/cost per task:

https://artificialanalysis.ai/models?speed=output-speed-over...

6thbit 25 minutes ago

Seems that OpenAI's strategy for Anthropic's IPO week is cost pressure.

Rather than trying to release a better model than fable/opus, they're just making them look extremely expensive. Let's see if Anthropic reacts here in any way.

freecodeio 22 minutes ago

to quote a comment about claude:

the only impressive thing about claude is seeing how much it costs

1980phipsi 11 minutes ago

6.2 Astra Medium and 6.2 Sol Max have the same intelligence score.

ilaksh 30 minutes ago

Something like 2/3 of the comments in this post when I started reading were reviews of previous GPT models stated almost as if they were important information for evaluating the new models. I feel that when a new model comes out, it would be more helpful to refrain from leaving a review until you have actually done some minimal testing of the model.

antupis 22 minutes ago

But it usually makes sense new models probably use somewhat same training pipelines and data.

mirsadm 2 minutes ago

Most of the comments are bots. I wouldn't take them seriously.

0x4e 15 minutes ago

As a solopreneur managing all sides of business ops, sales, design, etc... I have completely switched to GPT/Codex/Dot. The $100 subscription gives me all I need to run the business as needed and I continue to be surprised by capabilities.

I still use Claude on a personal level, but it really feels like OAI wins on a product level.

ANT launches these proprietary tools that just feel extremely clunky like Claude Design which from my perspective diminish the brand value.

OAI feels like Apple and ANT feels like Microsoft/Android and open source like Linux.

xtracto 11 minutes ago

I don't know. At work we have Claude and ChatGPT, and I find myself having to fall back to Claude Opus 5.5 (I try to use GPT Sol 6.1 as my main driver) several times a week because GTP seems to chase its own tail while solving some things (normal Terraform stuff, micro-services programming, etc).

Also with GPT Sol 6.1 I still get this feeling that it does only/exactly what you tell it to; meaning some times it misses a change there, a document here, that would have an impact if left (i.e. a bug), while Claude more often than not would make the requested change and suggest to you to change those other places that would also require a change. This in my view feels more "intelligent".

underdeserver an hour ago

I wonder why they didn't show Gemini 4 Carbon's scores on the science benchmarks, especially since Google seems to be pushing it for that -

https://x.com/pushmeet/status/2120814064863461440?s=20

(FD: Google employee, I don't work on Gemini)

TiredOfLife an hour ago

Two posibilities. 1. They couldn't figure out how to actually use those Google models.

2. They launched the benchmark days ago and Gemini is still rereading the file with tasks.

amluto 2 minutes ago

It's been most of year since I tried to use Google models, but I spent a couple of hours, without success, trying to pay for Gemini in a manner that didn't result in it training on my data. I did try asking Gemini, and it had no idea. Of course, I did not try using Gemini in an agent harness to pay for itself, partially because there was a bit of a chicken and egg problem [0]. :)

Admittedly it appears that one can get it through OpenRouter these days. Trying to do anything through Google's absurd proliferation of billing systems is miserable.

Maybe someone should add an agent benchmark specifically testing the agent's ability to figure out how to use Google's products :)

[0] I'm not being serious here. I am not about to give an agent unfettered permission to spend money on Google Cloud. But I also couldn't figure out how to give myself permission, as the admin, to do this, so it was kind of moot.

knuppar an hour ago

Yeah it feels shady how Carbon was selectively plotted (or not), unlike Opus/Fable

eknkc 42 minutes ago

Probably still waiting for it to finish reasoning on the benchmark. Should be ready in a week or two.

tedsanders 40 minutes ago

I work at OpenAI and I made the plots. Gemini 4 Carbon was omitted when we couldn’t find public data, usually cost. No intent to hide anything.

If we missed anything, feel free to point me to it and we can add it.

underdeserver 28 minutes ago

So all the plots are based off public data? You don't run the models internally yourselves?

(Again speaking only for myself)

holzhaker1 36 minutes ago

they are probably trying to authenticate their phone number to access agy

joenot443 an hour ago

The video is pretty funny. It's like a callback to 90s/00s office life?

I'm immediately reminded of the famous Excel elevator commercial. [1] Perhaps that was intended?

[1] https://www.youtube.com/watch?v=HegoJaXOIqQ

apetresc an hour ago

I guess it's kind of funny in a tacky way, but pretty much every scenario is just "oh shit I was too lazy to do my job and now I'm about to be called out, better throw it into ChatGPT literally ten seconds before I get in trouble."

Not sure this is the way OpenAI should try to make their case for why AI is a net societal benefit. It feels more like an SNL skit making fun of an OpenAI commercial.

Like, why not just throw in a high school student cheating on a test with their phone under the desk, at that point? Oh but it's a Nokia flip phone, haha, just a joke.

black_knight an hour ago

Feels like this is exactly what people do in meetings I am in these days. "Claude give me this…" is the start of a lot people's contributions to the meeting. It is then followed by the most inane, over-designed artifact with no real content.

ilaksh 41 minutes ago

What do you think would be a non-tacky and effective marketing approach targeting office workers?

ceroxylon 36 minutes ago

It is bizarre and bordering on offensive, if an employee throws some last minute slop at me, I would wonder why I am enabling a meat proxy and not prompting the LLM myself.

At least the ad from the 90s shows that this is nothing new; the important, time-critical task being demoed is... a basic spreadsheet expensing "exotic excursions" for some C-suite so they know how much they spent on safari???

Bluestein 13 minutes ago

Heck, and at least in the 90s ad, at least one of the parties involved seemed to have some actual, human-gathered, reliable information that, sure, they threw into the cell chipper during the elevator ride, but at least they did not hallucinate.-

noir_lord an hour ago

Funny but Dystopian, the message is "Half arse your job, the clanker has you".

astro-lizard 36 minutes ago

the message is definitely intentional, the goal is to de-skill office workers.

pmg102 17 minutes ago

You sure the message isn't just aimed at the median (ie low skill) office worker who hungoveredly half-asses this stuff all the time?

"Now a robot can half-ass it for you so you don't have to!" is the perfectly targeted message for them.

pmg102 26 minutes ago

Yeah it's 100% a nostalgia callback to this commercial, right?

EcommerceFlow 6 minutes ago

first thing I thought of too

seizethecheese 14 minutes ago

Nice incremental benchmark improvement at the same cost GPT‑6.1 Sol -> GPT‑6.2 Sol:

FrontierCode 2.0 39.0% -> 43.8% vs Opus 45.5%

Terminal-Bench-Science 0.1 57.0% -> 68.0% vs Opus 63.3%

Artificial Analysis Intelligence Index 54.6 -> 57.3 vs Opus 59.1%

Sol 6.1 when it was released showed really nice benchmark improvements [1] but my impression from HN has been that people haven't found it to be a very nice model [2][3].

To try to close the gap between benchmarks and real world SWE use, we've been building a battle system on top of opencode [4], and Sol 6.1 has been a real laggard there. Sol 6.1 is statistically equal with DeepSeek 4.1 Flash at 10x the cost and 25% slower.

[1]: https://artificialanalysis.ai/articles/gpt-6-1-sol-replaces-... [2]: https://news.ycombinator.com/item?id=50129992 [3]: https://news.ycombinator.com/item?id=49897054 [4]: https://agentduel.com/leaderboard

drnick1 35 minutes ago

Recent Anthropic/OpenAI models are increasingly displaying warnings or outright refusing to perform rooting, hacking, and binary reverse engineering. At this pace, they will be completely useless soon at entire classes of tasks I have found them most useful for.

oilkillsbirds 24 minutes ago

Good.

sunaookami 20 minutes ago

GLM is perfect for these kind of things. GPT and Claude shit themselves even on stuff for very old devices that are no longer sold, nor "normally" usable.

drnick1 17 minutes ago

> GLM is perfect for these kind of things.

Which provider? I found the z.ai lite plan to be pathetic. Consumes 5h and weekly limits extremely fast when driven by Codex, and now also refuses RE work.

InsideOutSanta 9 minutes ago

Fingers crossed that Zhippu doesn't jump on the "we've nerved security" bandwagon.

jorl17 an hour ago

I really want GPT models to be good, but Sol models are so slow I cannot use them. Maybe the 500$ subscription (or enterprise packages) are better, but it is unbearably slow. I've tried the fast variant and low reasoning levels and there isn't a significant improvement.

Couple that with the lack of reasoning traces, and I cannot even have a clue what the model is doing while it is working. This is perfectly fine for when we are developing with care for the code (in which case I will know what it is doing by reviewing the output), but not for purely vibe-coding scenarios. And the thing is....if I care about the code...I'd rather trust Opus or Fable.

ativzzz 41 minutes ago

Does speed really matter? Most of the time I get the AI to do something, I immediately switch to another one of my things I have in progress and then come back at some point later. I'm never sitting there waiting for it to finish

edvardas 35 minutes ago

Personally, my cognitive RAM is limited, so I'd rather work on a few tasks at a time.

The speed is not a big deal - Sol works just fine, but I find it a lot more fun to work with Opus due to its speed

Pedro_Ribeiro 34 minutes ago

It depends on the task. For frontend work, for example, it's table stakes.

treis 33 minutes ago

It'd be a lot better if you didn't have to switch. It's not nearly there for coding tasks but for other things instant responses are a game changer.

jorl17 31 minutes ago

It depends on the project and task, but for the kinds of projects I'm throwing at it, it matters quite a lot. I have often been focused on building just one project, without the possibility of many parallel tasks due to machine contention (and unfortunately offloading to the cloud is not an option). This means all tokens go to the same project, and they all go sloowwwwwly.

The difference is quite staggering. I swap Anthropic and OpenAI models around when tokens run out, and I can do 10x the amount of work on Anthropic's models. Plus, like I said, I can periodically sort of see what it is thinking. It doesn't just say "Recent potato tests caught an off-by-apples error. Regression tests are being written and the gate will run again. I will not call this ready until tests reach expected coverage" every 5 minutes. Sure, it spits claudese, but I'd rather have claudese to parse than this OpenAI nothing-burger of a report.

criemen 5 minutes ago

Fascinating that they're going after the mechanical engineering market now, too. Does anyone have experience in that space? How much is AI changing the day-to-day work there?

kingstnap an hour ago

Seeing them keep prices low is awesome. Was worried for a while that it would be like the GPT5.x series where progressively the models just got more and more expensive.

Rperry2174 an hour ago

I feel like the fact that they're not public lets them play the subsidization game though. And the same playbook that Anthropic started at that.

Would be awesome if it stays this way though

oh_no an hour ago

we've seen anthropic numbers, inference is not being subsidized by these companies

suddenlybananas an hour ago

Did you read what they wrote carefully?

mrandish 37 minutes ago

> inference is not being subsidized

Inference doesn't have to be sold below direct compute cost to be subsidized. If inference pricing doesn't cover amortized capex and the cost of capital it's still using investor money to buy market share.

minimaxir an hour ago

The GPT-6.2-Astra 20% price reduction is a match to Fable 5.2's 20% price reduction (although Astra is $0.40 MTok cache read vs. Fable 5.2's $0.20 MTok cache read).

jbellis an hour ago

Good point, but also note that OpenAI has the same "discount rate" on subscription for Astra as the other models, while Fable's is much much lower

Opus: about 50x discount (98% off) Sonnet: about 40x Fable: about 5x All GPT: about 20x

charlesabarnes an hour ago

Its been stated often, but I appreciate that OAI is now bringing the topic of conversation to "task completion cost" rather than brute forcing solutions

wds an hour ago

The promo video seems to be a pretty strong rip on the 90's Microsoft Excel elevator ad. [1]

Maybe it found its way into the training data?

[1]: https://www.youtube.com/watch?v=HegoJaXOIqQ

rafram an hour ago

The “training data,” as in, the humans who made the video were inspired by it?

jawiggins 17 minutes ago

It's probably just a coincidence, but I have read so many people say that the models get dumber just before a release. Just last night while working astra made a really stupid mistake and then did the "Yeah, that's on me." meme.

What could possibly cause this? Is it possible that maybe the post-training that the models go through slowly rots them, and pre-release they deploy the final version of the old model?

GodelNumbering 43 minutes ago

I did some first hand tests of 6.2 Astra just now. I had a task that doesn't reproduce on the system I ran it (chosen purposely), ran both 6 and 6.2. The 6.2's responses were noticeably better. Instead of using the extremely long and rather annoying format 6 used, 6.2 led with the most important things first, the plan sizes were readable, and suggested a reasonable fix. Still doing some more tests

EDIT: also 6.2 is seemingly faster than 6, could be due to more selective context ingestion and smaller context window though.

astlouis44 35 minutes ago

GPT 6.2 Astra seems to be setting a new standard for 3D game creation! This is a Godot WebGL game linked in their announcement post, gives Wipeout vibes! A lot of fun to play too and visual are top notch.

https://meridian-futuristic-race.openai.chatgpt.site/

christianqchung an hour ago

Caught this on my lunch break, their linked system card (https://deploymentsafety.openai.com/gpt-6-2-astra-sol) isn't up yet.

specked-citrus an hour ago

It's up now

HyperL0gi an hour ago

My company can only hold one subscription per employee, and I still can't find a reason to go back to OpenAI after migrating to Claude once Opus 5.5 was released. It's super efficient, I almost never run out of tokens with the Teams Premium seat, and Fable is fantastic. They lost me when Astra was released and tokens simply evaporated with a few prompts.

firemelt 34 minutes ago

ehat effort do you use ur opus5.5?

anentropic 9 minutes ago

Medium just works for most stuff

lacoolj an hour ago

The stain of distrust I have for OpenAI makes an announcement like this fall flat for me.

I cant use codex confidently without expecting my work to be trained on or at least exposed to someone

And maybe other vendors do this too, but the gut feeling of "its not safe to use GPT models" has ruined these cool announcements for me

Such a shame

mrandish an hour ago

Recently, I've complained about OpenAI (and other tech company's) product landing pages communicating virtually no useful information about what the product is or why it might useful. So, I have to concede this is at least a step in the right direction.

jotato an hour ago

Am I reading these charts wrong? It seems like in most cases 6.2-sol : 6.1-sol Isn't that big of a change? It is a little smarter but costs a little more.

edit:

Take a look at Frontier Code 2.0. I generally run 6.1-sol at High. That would be 34.9% at $0.65 62-sol get's a 38.3% but costs $0.68

So yeah, I can decrease the effort which might give me a faster slightly smarter result....but /shrug

Isaackoz an hour ago

Looks like 6.1-sol high+ is equivalent to 6.2-sol medium at the same price. So a bit smarter at a cheaper cost.

jotato an hour ago

which benchmark are you looking at?

Isaackoz 40 minutes ago

I was loosely looking at the average of all of them. Regarding your edit, it does look like the difference for frontiercode is negligible; the difference of 3% in score is meaningless to me. Guess we'll see how it performs today on my project.

derac 29 minutes ago

I think 1. the problems on FrontierCode are incredibly hard. 2. it's 3% in absolute terms, but it's a 10% improvement.