Rendered at 12:46:00 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
wincy 17 hours ago [-]
Just tried it out for a prod issue was experiencing. Claude never does this sort of thing, I had it write an update statement after doing some troubleshooting, and I said “okay let’s write this in a transaction with a rollback” and GPT-5.5 gave me the old “okay,
BEGIN TRAN;
-- put the query here
commit;
I feel like I haven’t had to prod a model to actually do what I told it to in awhile so that was a shock. I guess that it does use fewer tokens that way, just annoying when I’m paying for the “cutting edge” model to have it be lazy on me like that.
This is in Cursor the model popped up and so I tried it out from the model selector.
XCSme 17 hours ago [-]
I feel like the last 2-3 generations of models (after gpt-5.3-codex) didn't really improve much, just changed stuff around and making different tradeoffs.
pixel_popping 17 hours ago [-]
I disagree, it improved enormously especially at staying consistent for long-tasks, I have a task running for 32 days (400M+ tokens) via Codex and that's only since gpt-5.4
ericpauley 17 hours ago [-]
Has that task accomplished anything yet?
codemog 17 hours ago [-]
I think the OP is in for a rude surprise when the task is “finished”.
hagbard_c 16 hours ago [-]
It will go somewhat like this:
“You're really not going to like it," observed Codex.
"Tell us!"
"All right, said Codex. "The answer to your Great Question..."
"Yes...!"
"Is..." said Codex, and paused.
"Yes...!"
"Is..."
"Yes...!!!...?"
"Forty-two," said Codex, with infinite majesty and calm.
pixel_popping 16 hours ago [-]
I bet you've asked Codex for that joke :p
xp84 17 hours ago [-]
Too soon to tell, give it a billion tokens before we make up our minds
pixel_popping 16 hours ago [-]
Oh boy, you are far from what it requires, we are probably talking 3B+, but note that this is just codex, obviously codex is also doing automatic adversarial with the regular zoo (gemini-3.1-pro-preview, opus-4.6/4.7, gpt-5.3-codex, minimax-2.7, glm-5.1, mimo-2 (now 2.5) and so-on, you get the gist) :)
fl4regun 16 hours ago [-]
what is that task doing???
owebmaster 2 hours ago [-]
Interesting is that they had the opportunity to explain but decided that hyping it more made more sense. 3 billion tokens!!1!
anonym00se1 11 hours ago [-]
The correction question is: what isn't that task doing?
elAhmo 13 hours ago [-]
It made Sam richer.
SecretDreams 17 hours ago [-]
Kept the OP employed for a full extra month at their high AI metric firm, hopefully.
pixel_popping 15 hours ago [-]
Just making Jensen proud is all.
lowdude 16 hours ago [-]
That’s actually crazy, what kind of task is that? And is that a recurring kind of task like some analysis, or coding related?
pixel_popping 16 hours ago [-]
Coding (along with docs, tests obviously), rewriting a huge chunk of the KVM hypervisor (in Kernel 7, started in the -rc2) and KSM and other modules, can't say too much about it yet (might do an announcement in coming weeks). The coding is automated but the plan took days of manual arguing (with all models possible) prior (while doing other things during waiting times as I currently manage 70 repos for an upcoming release of our Beta).
I think users really underestimate the capabilities of "AI" when using the right tooling/combinations of models and procedures (and loops), that's talking with 2 decades of dev behind me, genuinely I'm not on phase with people saying it produces slop of any kind, at this stage, it's mostly the fault of the prompter (or the prompter not having enough tokens to do mass adversarial), but clearly, I can genuinely state that the code produced is overall the SAME quality as I would by being extremely meticulous.
I'm like a bot following 30+ threads concurrently, sometimes it's fun, sometimes it feels like playing casino, sometimes it's boring, but this is truly an insane era if you have the funding for it, obviously we stack many MANY accounts in rotation 24/7, equivalent in API cost by myself is about 100K$+ (a month) but we pay only a fraction of that cost thanks to the plans.
PS: I have 8 monitors in front of me to manage all that (portable monitors stacked together).
Urahandystar 16 hours ago [-]
Please do an update when you're ready, this sounds like madness to me so I'd love to see what the output is. Whatever it is I have to know.
owebmaster 2 hours ago [-]
Typical AI psychosis. They might notice it soon or stay in this condition for months.
AlexCoventry 12 hours ago [-]
Is it hitting intermediate milestones with solid pre-written and human-reviewed acceptance tests? If not, sounds like a very risky commitment.
ericreg92 14 hours ago [-]
Please do a post about this (though I realize that takes time). This sounds amazing. I have always dreamed of doing this too but just don't have the budget.
stirfish 10 hours ago [-]
Specifically, write a post about this and do not have Claude write a post about this.
7thpower 14 hours ago [-]
I have yet to talk to someone who is taking this approach and doesn’t end up with a dumpster fire, but here is to hoping this time is different.
Hope it works and you post about it.
Culonavirus 12 hours ago [-]
I hope it doesn't work and they don't post about it.
ziml77 12 hours ago [-]
It's just too bad the subsidized costs mean they won't actually feel any real punishment for their failure. Like normally time wasted on its own is enough of a punishment for making a poor decision, but they're not even doing anything themselves here!
jamwil 14 hours ago [-]
I’m vague on a specific reason for this feeling because there are a few to choose from and no one overpowers the other, but the emotion that comes to mind when I read this is disgust. As a society I feel we will look back on the subsidized opulence of this moment with total and utter contempt.
deaux 4 hours ago [-]
I know exactly the feeling you mean. I get a much stronger feeling of that when I talk with friends who frequently take a plane for a 250 mile trip which has a world-class comfortable high-speed train connection with very frequent trains, each taking less than 3 hours. I'm sure you have friends who would do this in this situation - do you feel the same disgust when you hear them talking about such choices?
I still haven't seen a single person who actually cares about the environment and has willingly made significant sacrifices for it, who clamors about the environmental cost of AI. Every time I see someone do it it's someone who never cared about this before, and still doesn't really. Who buys plenty of new clothes and furniture, loves a good burger, has the latest iPhone, flies 4 times per year.
Maybe you're the unicorn in which case fair enough, you've earned the right to feel disgusted.
owebmaster 2 hours ago [-]
There's no opulence in spending tokens for entertainment. Vibecoding your own game is the new viral game.
holmesworcester 14 hours ago [-]
Or nostalgia for simpler times
jamwil 14 hours ago [-]
That as well. But everyone reading GP’s posts knows in their bones that it’s unsustainable. It’s economically unsustainable and environmentally unsustainable, and in that context it strikes me as pure hoarding behaviour. Taking as much as they can for themselves before the house of cards crashes down.
I have no sympathy for OpenAI or Anthropic as corporations, but if these are the new tools of the trade, then platform abuse like GP is bragging about serves only to destroy the livelihoods of the rest of us who are content to use our fair share.
There’s no such thing as a free lunch, and the bill always comes at the end.
owebmaster 2 hours ago [-]
> (might do an announcement in coming weeks).
Don't be surprised if/when people ignore your AI slop
r_lee 16 hours ago [-]
...what? what kind of a task are you running?
ninkendo 10 hours ago [-]
Sorry if I’m not getting it, but what was wrong exactly? Is the issue that it merely put “-- put the query here” in the reply, instead of repeating it again?
If so, I’m not sure I’d even consider that a problem. If the goal is for it to give you a query to run, and you ask it “let’s do it in a transaction”, it’s a reasonable thing for it to simply inform you, “yeah you can just type begin first” since it’s assuming you’re going to be pasting the query in anyway. And yeah, it does use fewer tokens, assuming the query was long. Similar to how, if it gave me a command to run, and I say “I’m getting a permission denied”, it would be reasonable for it to say “yeah do it as root, put sudo before the command”, and it’s IMO reasonable if it didn’t repeat the whole thing verbatim just with the word “sudo” first.
But if the context was that you actually expected it to run the query for you, and instead it just said “here, you run it”, then yeah that’s lazy and I’d understand the shock.
endymi0n 16 hours ago [-]
OpenAI is the first company that has reached a level of intelligence so high, the model has finally become smart enough to make YOU do all the work. Emergent behavior in action.
All earnesty aside, OpenAI’s oddly specific singular focus on “intelligence per token” (also in the benchmarks) that literally noone else pushes so hard eerily reminds me of Apple’s Macbook anorexia era pre-M1. One metric to chase at the cost of literally anything else. GPT-5.3+ are some of the smartest models out there and could be a pleasure to work with, if they weren’t lazy bastards to the point of being completely infuriating.
syspec 17 hours ago [-]
Can't tell if above is good or bad.
wincy 10 hours ago [-]
I mean, I was doing triage, so wanted an immediate fix. The actual issue is we’re getting some exploding complexity when double checking the action the API is taking is valid in the data. So that needs to be refactored. I suppose it reduces token usage, but Claude Opus will happily do exactly what I want it to.
hbn 16 hours ago [-]
GPT-5.5 shatters benchmarks for amount of faith it puts in the user.
croemer 45 minutes ago [-]
It refuses to write bioinformatics code that involves analysis of SARS-CoV-2. Even when it's totally obvious I'm not trying to do any bioengineering of any sorts. Totally harmless stuff I'm doing and I just get rejected.
I know it's only on a single benchmark, but I dont understand how it can be so bad...
goldenarm 16 hours ago [-]
gemma4-e4b is 50% better than gemma4-26b in your benchmark, something's wrong
guilamu 16 hours ago [-]
Yes those two models were tested on my own PC (local inference using my own CPU/GPU). So something my be bugged on my setup. gemma4-26b should be far better than gemma4-e4b.
data-ottawa 21 minutes ago [-]
The early quants for Gemma4 26b had issues and needed to be updated, might be worth checking
embedding-shape 16 hours ago [-]
Sounds like maybe using worse quantization on the bigger model? Quantization matters a lot for the quality, basically anything below Q8 is borderline unusable. If it isn't specified in a benchmark already it probably should.
ac29 17 hours ago [-]
Your benchmark has Opus 4.7 performing significantly worse than Sonnet 4.6. Even if true on your benchmark, that is not representative of the overall performance of the models.
guilamu 16 hours ago [-]
Yes Opus 4.7 fast (no reasoning) did a worst job than Sonnet 4.6 high (with reasoning) according to Gemini 3.1 Pro evaluation.
ac29 16 hours ago [-]
Your table doesn't indicate reasoning vs non-reasoning, or reasoning level
guilamu 16 hours ago [-]
When nothing is noted it's max reasoning (xhigh in copilot chat in vscode if available).
The models not availble on copilot were tested through opencode (max reasoning) and deepseek v4 was tested through Cline (with max reasoning too).
mosselman 16 hours ago [-]
You even traveled in time to deliver us this benchmark.
I really like this benchmarking. Have you evaluated the judge benchmark somehow? I'd love to setup my own similar benchmark.
guilamu 16 hours ago [-]
Haha, just fixed the date!
I haven't evaluated the judge benchmark. You have everything needed in the repo to do so though, so be my guest. It took me a bit of time to put all this together and won't have much more time to dedicate to it before a couple of weeks.
BTW, if you explore the repo, sorry for all the French files...
DrProtic 16 hours ago [-]
Seems like benchmark for how good a model is for vibe coding.
Your prompt is extremely slim yet you score it on a bunch of features.
guilamu 16 hours ago [-]
Yes, the prompt is slim by design. I might be wrong, but the point was to see what the model can do "on it's own".
That’s the thing, not everyone wants and values the model based on that. But I guess it works for you, and that benchmark achieves it.
I personally develop with very detailed spec, and I don’t want nothing more and nothing less compared to the spec.
I found 5.4/5.5 much better at following spec while Opus makes some things up, which aligns with your benchmark but that makes 5.4/5.5 better for me while worse for you.
guilamu 15 hours ago [-]
Yeah as I said this a benchmark for my usecase only, a single use case, which is obvisouly not representative of everybody's needs.
What strike me as very strange though is that 0 model were able to just use the search input already present in GravitYForms forms list page and all created a second input.
Also, I know it's not in the prompt, but adding a ctrl+f shortcut to a search input? Is that that crazy? I don't know.
gizmodo59 12 hours ago [-]
[dead]
Topfi 15 hours ago [-]
Pricing by context length:
Input: $5/M tokens at <=272K, $10/M tokens above 272K.
Output: $30/M tokens at <=272K, $45/M tokens above 272K.
Cache read: $0.50/M tokens at <=272K, $1/M tokens above 272K.
Significantly more expensive than Opus 4.7 beyond 272K and at least in my tasks, I haven't seen the model that much more token efficient, certainly not to such a degree that it'd compensate this difference. GPT-5.4 had a solid context window at 400k with reliable compaction, both appear somewhat regressed, though still to early to truly say whether compaction is less reliable. Also, I have found frontend output to still skew towards that one very distinct, easily noticeable, card laden, bluesy hue overindulged template that made me skeptical of Horizon Alpha/Beta pre GPT-5s release. Ended up doing amazing at the time for task adherence, which made it very useful for me outside that one major deficit. The fact that GPT-5.5 is still so restricted in that area is weird considering it's supposed to be an entirely new foundation.
sigmoid10 18 hours ago [-]
Huh. Yesterday they said:
>API deployments require different safeguards and we are working closely with partners and customers on the safety and security requirements for serving it at scale.
And now this. I guess one day counts as "very soon." But I wonder what that meant for these safeguards and security requirements.
FINDarkside 18 hours ago [-]
When stuff is delayed due to "safeguards" it just means they don't think they have the compute to release it right now.
simonw 18 hours ago [-]
I wonder if the fact that GPT-5.5 was already available in their Codex-specific API which they had explicitly told people they were allowed to use for other purposes - https://simonwillison.net/2026/Apr/23/gpt-5-5/#the-openclaw-... - accelerated this release!
embedding-shape 18 hours ago [-]
The same person who've mercilessly lied about safety is still running the company, so not sure why anyone would expect any different from them moving forward. Previous example:
> In 2023, the company was preparing to release its GPT-4 Turbo model. As Sutskever details in the memos, Altman apparently told Murati that the model didn’t need safety approval, citing the company’s general counsel, Jason Kwon. But when she asked Kwon, over Slack, he replied, “ugh . . . confused where sam got that impression.”
In the only one that feels that OpenAI has bots/commenters on payroll on all this kind of news downplaying Claude and stating how much better Codex is?
There is too much and there are too many, and some of their takes don’t fly if you use Claude daily.
AlexCoventry 11 hours ago [-]
Yeah, it's eerie, same with how everyone seems to have forgotten that OpenAI betrayed democracy by committing to work on unsupervised autonomous weapons and domestic mass surveillance.
int3trap 8 hours ago [-]
Honestly I find comments like yours much more eerie. By all accounts they never agreed to any of that but you say it with such confidence like it's a fact.
AlexCoventry 6 hours ago [-]
The Trump administration's handling of Anthropic showed that regardless of what the contract or the law says or means, they will severely penalize any vendor who refuses their demands. And OpenAI stepped right into that relationship immediately after the administration showed that. So either they were signing up for a supply-chain risk designation and whatever other punishments the Trump administration dreams up, or they're complying.
If this sounds crazy to you, though, I'd like to know, and understand why. I miss ChatGPT/Codex.
psteinweber 2 hours ago [-]
I find that very obvious too. It started (visibly) shortly after the Opus 4.6 hype.
Aboutplants 12 hours ago [-]
Of course they do. As do all of the other companies pushing their product these days.
Enterprise user here and still seeing only 5.4.
Yesterday's announcement said that it will take a few hours to roll out to everybody. OpenAI needs better GTM to set the right expectations.
neosat 17 hours ago [-]
Just refreshed and see 5.5 now - yay! Love the speedy resolution ;) Thanks folks, I'll complain faster next time....
17 hours ago [-]
gertlabs 13 hours ago [-]
Comprehensive coding reasoning benchmark results for GPT 5.5 with max reasoning are up at https://gertlabs.com/
Live decision and heavier agentic evals will continue being uploaded for 24 hours but I don't expect its leaderboard position to change at this point.
GPT 5.5 is the most intelligent public model. And significantly faster than its predecessor.
czk 17 hours ago [-]
API page lists the knowledge cutoff as Dec 01, 2025 but when prompting the model it says June 2024.
Knowledge cutoff: 2024-06
Current date: 2026-04-24
You are an AI assistant accessed via an API.
BeetleB 17 hours ago [-]
I don't know why this keeps coming up. This has always been the least reliable way to know the cutoff date (and indeed, it may well have been trained on sites with comments like these!)
Just ask it about an event that happened shortly before Dec 1, 2025. Sporting event, preferably.
czk 17 hours ago [-]
the model obviously knows things after the reported date but its just curious that it reports that date consistently
could be they do it intentionally to encourage more tool calls/searches or for tuning reasons
jumploops 15 hours ago [-]
[dead]
htrp 17 hours ago [-]
Can you really believe things that the model says? (A lot of prior model api pages say knowledge cutoffs of June 2024, maybe the model picks that up?)
czk 17 hours ago [-]
you cant but its pretty reproducible across api and codex and other agents so i just thought it was odd. full text it gives:
Knowledge cutoff: 2024-06
Current date: 2026-04-24
You are an AI assistant accessed via an API.
# Desired oververbosity for the final answer (not analysis): 5
An oververbosity of 1 means the model should respond using only the minimal content necessary to satisfy the request, using
concise phrasing and avoiding extra detail or explanation."
An oververbosity of 10 means the model should provide maximally detailed, thorough responses with context, explanations, and
possibly multiple examples."
The desired oververbosity should be treated only as a *default*. Defer to any user or developer requirements regarding
response length, if present.
17 hours ago [-]
bakugo 17 hours ago [-]
Models don't know what their cutoff dates are unless told via a system prompt.
The proper way to figure out the real cutoff date is to ask the model about things that did not exist or did not happen before the date in question.
A few quick tests suggest 5.5's general knowledge cutoff is still around early 2025.
czk 17 hours ago [-]
i wonder if they put an older cutoff date into the prompt intentionally so that when asked on more current events it leans towards tool calls / web searches for tuning
ssl-3 16 hours ago [-]
I wonder if the cutoff date is the result of so many people posting about the date over time and poisoning the data. "Dead cutoff date theory," perhaps.
Whatever it is, the cutoff date reporting discrepancy isn't new. Back when Musk was making headlines about buying/not buying Twitter, I was able to find recent-ish related news that was published well after the bot's stated cutoff date.
ChatGPT was not yet browsing/searching/using the web at that point. That tool didn't come for another year or so.
MallocVoidstar 17 hours ago [-]
OpenAI does tell the model the current date via API, so it's odd for them not to also tell the model its cutoff
soco 17 hours ago [-]
Stupid question: wouldn't it then search the web for that event?
bakugo 17 hours ago [-]
If you have web search enabled, sure. But if you're testing on the API, you can just not enable it.
swyx 17 hours ago [-]
can u test it on say who won the 2024 US election
ghurtado 17 hours ago [-]
I can't really think of a less reliable test for anything at all than making a random guess as to something that had about 50/50 odds to begin with
Easiest Turing test ever...
himata4113 17 hours ago [-]
ask it 10 times.
pixel_popping 17 hours ago [-]
MASSIVE ADVERSARIAL x50
WarmWash 17 hours ago [-]
Usually the labs do some kind of post training on major events so the model isn't totally lost.
A better test is something like "what is the latest version of NumPy?"
bakugo 17 hours ago [-]
That sort of test isn't super reliable either, in my experience.
You're probably better off asking something like "what are the most notable changes in version X of NumPy?" and repeating until you find the version at which it says "I don't know" or hallucinates.
czk 17 hours ago [-]
with thinking off and tools disabled:
Donald Trump won the 2024 U.S. presidential election.
redsocksfan45 17 hours ago [-]
[dead]
jumploops 15 hours ago [-]
[dead]
robertwt7 15 hours ago [-]
Gpt 5.5 combined with codex is really good. I actually have no doubt whenever I asked questions, plan, or implement a code with it. With opus 4.7, I have to keep double checking because it doesnt follow the CLAUDE.md instruction, it hallucinates a lot, by default it makes things up when it can’t find the answer to something. Its crazy how quickly people are saying that OpenAI is left behind last year when they declared code red and look at where we are now
zerof1l 16 hours ago [-]
I don't see any meaningful performance improvements in those paid models anymore.
They all roughly produce junior developer-level code, continue to have mental breakdowns in their “thinking” stage, occasionally hallucinate things, delete pieces of code/docs they don’t understand or don’t like, use 1.5 times the necessary words to explain things when generating docs and so on.
I'm now testing "avoid sycophancy, keep details short and focus on the facts" in my AGENTS.md files.
podnami 16 hours ago [-]
This is snark. Since when has a junior level dev managed to debug and deploy say a cloudformation stack and follow up with notes under 3 minutes?
gjsman-1000 15 hours ago [-]
Heard this analogy elsewhere, but worth repeating:
AI is like having the greatest developer who ever lived, but she is always on 4 beers.
nathan-hello 15 hours ago [-]
personifying ai is incredibly cringe no matter how weird your comparison is
jubilanti 11 hours ago [-]
It's an analogy.
nathan-hello 2 minutes ago [-]
that’s the personification i’m referring to, yes. incredibly weird.
gjsman-1000 15 hours ago [-]
Imagine a drunk developer. Sparks of brilliance while missing obvious trees.
nathan-hello 34 seconds ago [-]
weird.
nimchimpsky 15 hours ago [-]
[dead]
ftonon 17 hours ago [-]
Looks like the default config in the chat is instant 5.3, it only uses the 5.5 on the thinking variant
bnm04 16 hours ago [-]
They moved a few months ago to have separate instant and thinking models. 5.3 is the latest instant, and 5.5 is a reasoning model.
QuadrupleA 16 hours ago [-]
Exactly double the cost of GPT 5.4 - $5 per MTok input, $0.50 cached, $30 output.
All the AI players definitely seem to be trying to claw more money out of their users at the moment.
languid-photic 16 hours ago [-]
It's 2x/token, but for default reasoning we've found GPT-5.5 uses fewer tokens overall, so net cheaper on median. [1]
(Note, that stops being true at higher reasoning levels, where our observed total cost goes up ~2-3x.)
This is where the emigration to Chinese providers begins.
throw03172019 18 hours ago [-]
Faster than anticipated because of Deepseek release?
XCSme 17 hours ago [-]
Doubt it, DeepSeek v4 is quite underwhelming.
swyx 17 hours ago [-]
more like they wanted to release it yesterday but merely had some last min flags they wanted to hold off for
14 hours ago [-]
Jhonwilson 17 hours ago [-]
ok not bad
m3kw9 17 hours ago [-]
Maybe but no one serious is using deepseek
brianbest101 18 hours ago [-]
[dead]
pants2 18 hours ago [-]
Is anyone here actually using pro models through the API? I'd be very curious what the use-case is.
chadash 18 hours ago [-]
Yes. High value work where cost (mostly) doesn't matter. For example, if I need to look over a legal doc for possible mistakes (part of a workflow i have), it doesn't matter (in my case) whether it costs $0.01 or $10.00, since it's a somewhat infrequent event. So i'll pay $9.99 more, even if the model is only slightly better.
bogtog 17 hours ago [-]
I'm surprised I never heard people talking about using -Pro variants, even though their rates ($125-175/M?) aren't drastically larger than old Opus ($75/M), which people seemed to use
freedomben 18 hours ago [-]
Indeed, even just Terms of Service and Privacy Policy work. Infrequent enough that cost isn't an issue, but model quality absolutely is
ComputerGuru 18 hours ago [-]
Yes? The same reason you would use it via the tooling.
A very expensive model for API usage. Fine in codex I think.
gigatexal 17 hours ago [-]
what's the real world comparison to opus 4.7 fellow coders?
Sembiance 15 hours ago [-]
I gave 4.6, 4.7 and GPT 5.5 the same prompt and task to reverse engineer a collection of sample vector files from an obscure Amiga CAD program and create a detailed txt specification and a python converter that converts to SVG and produce a report so I can visually verify.
4.6 did very well. 90% perfect on first try, got to 100% with just a few followups.
4.7 failed horribly. First produced garbage output and claimed it was done, admitted it did that when called out, proceeded to work at it a lot longer and then IT GAVE UP.
GPT 5.5 codex was shockingly good. Achieved 90% perfect on first try in about a fourth of the time. Got to 100% faster and with fewer follow-ups.
I’m impressed.
gigatexal 5 hours ago [-]
Interesting that 4.7 failed like that. Seems 5.5 is impressive but is oh so expensive.
Would be interesting if you ran your same test with Deepseek v4 and some of the other Chinese models.
pillefitz 17 hours ago [-]
Please consider the ethical aspects of giving money to OpenAI versus alternatives.
AlexCoventry 11 hours ago [-]
You need to be more specific. OpenAI's commitment to assist the Trump administration with domestic mass surveillance seems to have been largely memory-holed.
pillefitz 9 hours ago [-]
You're right, unfortunately. How naive of me to think that at least the HN audience would care.
refulgentis 16 hours ago [-]
I'm absolutely stunned by what I've seen from 5.5. I thought it'd be a nothingburger and ~= Opus.
Gave it two very long-running problems I haven't had the courage to work on in the last 2.5 years, solved each within an hour.
- An incremental streaming JSON decoder that can optionally take a list of keys to stop decoding after. 1800 LOC about 30 minutes later, and now my local-first apps first sync time is 0.8s instead of 75s when there's 1.5 GB of data locally.
- Flutter Web can compile to WASM and then render via Skia WASM. I've been getting odd crashes during rapid animation for months. In an hour, it got Skia WASM checked out, building locally, a Flutter test script, and root caused the issue to text shadows and font glyphs (technically, not solved yet, I want to get to the point we have Skia / Flutter patch(es))
If you told me a week ago that an LLM could do either of these, without heavy guidance, I'd be stunned. And I regularly push them to limits, ex. one of Opus' last projects was a tolerant JSON decoder, and it ended up being 8% faster than the one built-in to Dart/Flutter, which has plenty of love and attention. (we're cheating a little, that's why it's faster. TL;DR: LLMs will emit control characters in JSON and that's fine for me, treating them as fine means file edit error rates go from ~2% to 0%)
I just wish it was cheaper, but, don't we all...
Jhonwilson 17 hours ago [-]
that is great news
willj 15 hours ago [-]
[dead]
benjiro3000 17 hours ago [-]
[dead]
woohin 15 hours ago [-]
[dead]
theo_park87 17 hours ago [-]
[dead]
XCSme 16 hours ago [-]
GPT 5.5 is close to Opus 4.7, but at 7x the cost[0]...
Either Opus 4.7 miscounts reasoning tokens, or it's A LOT more efficient than GPT 5.5
I thought they made GPT 5.5 more token efficient than 5.4, but it uses 2x the reasoning tokens.
Very bad habit these safeguards. These "safety" filters are counter-productive and even can be dangerous.
In my place for example, a lot of doctors are using ChatGPT both to search diagnosis and communicate with non-English speaking patients.
Even yourself, when you want to learn about one disease, about some real-world threats, some statistics, self-defense techniques, etc.
Otherwise it's like blocking Wikipedia for the reason that using that knowledge you can do harmful stuff or read things that may change your mind.
Freedom to read about things is good.
NicuCalcea 17 hours ago [-]
> a lot of doctors are using ChatGPT both to search diagnosis and communicate with non-English speaking patients
I think that's the problem. Who's going to claim responsibility when ChatGPT hallucinates or mistranslates a patient's diagnosis and they die? For OpenAI, this would at best be a PR nightmare, so that's why they have safeguards.
rvnx 16 hours ago [-]
Adults bear responsibility for choices about their own lives. In fact, the more educated they are, the better choices they can make.
A doctor who gets refused by ChatGPT doesn't stop needing to communicate with the patient; they fall back to a worse option (Google Translate, a family member interpreting, guessing). Refusal isn't safety, it's liability-shifting dressed up as safety.
If there's no doctor, no interpreter, no pharmacist, just a person with a sick kid and a phone, then "refuse and redirect to a professional" is advice from a world that doesn't exist for them. The refusal doesn't send them to a better option; there is no better option, it's a large majority of people on this planet.
Hell is paved of good intentions, but open-education and unlimited access to knowledge is very good.
It doesn't change the human nature of some people, bad people stay bad, good people stay good.
About PR, they're optimizing for not being the named defendant in a lawsuit or the subject of a bad news cycle, it's self-interest wearing benevolence as a costume.
This is because harms from answering are punishable (bad PR, unhappy advertisers, unhappy investors, unhappy politicians / dictators, unhappy lobbies, unhappy army, etc); but harms from refusing are invisible and unpunished.
14 hours ago [-]
NicuCalcea 15 hours ago [-]
> A doctor who gets refused by ChatGPT doesn't stop needing to communicate with the patient; they fall back to a worse option
I think AI proves the contrary. There are plenty of examples of things that are getting worse because of technological advancement, particularly AI. Software quality, writing, online discourse, misinformation have all suffered over the last few years. I truly believe the internet is a worse place than it was 5 years ago, and I can't imagine bringing that to medicine would work out differently.
The medical system shouldn't rely on falling back to crappy workarounds, it should aspire to build the best system it reasonably can.
hellohello2 17 hours ago [-]
The doctor would be responsible.
I had a choice better a doctor that used AI or not, I would much prefer one that did...
NicuCalcea 17 hours ago [-]
The doctor would be responsible for the accuracy of their translation tool, something they can't verify but you expect them to use?
hellohello2 12 hours ago [-]
I was answering for hallucinations, not really for translation. Re-reading your initial post I do agree with what you are saying (i.e. you are explaining why OpenAI is looking to avoid a PR nightmare). What I meant to express is that I would personally trust doctors to use these tools as best they can to provide care.
lacunary 15 hours ago [-]
"what you see is all there is." it's generally much easier to verify something you've been made aware of than it is to know of it in the first place (and still verify it.)
rvnx 14 hours ago [-]
The irony is that licensed interpreters / translators usually perform worse than AI.
Only the liability shifts from OpenAI to them.
Furthermore, where the alternative to a licensed professional was nothing, or a random untrained person or a weak professional, then it's harming the user on the pretext of protecting him.
(like in the other mentioned contexts).
rvnx 16 hours ago [-]
What's the alternative then ?
-> You are in China, you go to emergency, nobody speaks your language
Move hands ? DeepSeek is better than using hands, even Baidu Translate, ChatGPT or whatever you find.
Other solutions are theoretically nice on paper but almost delusional.
An imperfect solution is better than no solution.
==
Similarly, a deaf-person is theorically better with a certified interpreter that can talk with the hands, but they may prefer voice-recognition software or AI tools.
(or... talking with hands is more confusing and annoying or less understandable for them).
Of course ChatGPT transcription can have issues, but that's the difference between the real-world and Silicon Valley's disconnected lawyers world.
==
If ChatGPT says: "sorry I won't be able, please go to see a licensed interpreter, good luck!" then it's just OpenAI trying to save their asses, at your risk/expense.
If you have a choice, you can make the choice, and you can double-check what is said. In other cases, you have no choice, nothing to check, only problems but no hints of solutions.
This is why openness is important.
NicuCalcea 15 hours ago [-]
When I registered with my GP in the UK, they asked me whether I would need an interpreter and what language. They then provide professional interpreters.
BEGIN TRAN;
-- put the query here
commit;
I feel like I haven’t had to prod a model to actually do what I told it to in awhile so that was a shock. I guess that it does use fewer tokens that way, just annoying when I’m paying for the “cutting edge” model to have it be lazy on me like that.
This is in Cursor the model popped up and so I tried it out from the model selector.
“You're really not going to like it," observed Codex.
"Tell us!"
"All right, said Codex. "The answer to your Great Question..."
"Yes...!"
"Is..." said Codex, and paused.
"Yes...!"
"Is..."
"Yes...!!!...?"
"Forty-two," said Codex, with infinite majesty and calm.
I think users really underestimate the capabilities of "AI" when using the right tooling/combinations of models and procedures (and loops), that's talking with 2 decades of dev behind me, genuinely I'm not on phase with people saying it produces slop of any kind, at this stage, it's mostly the fault of the prompter (or the prompter not having enough tokens to do mass adversarial), but clearly, I can genuinely state that the code produced is overall the SAME quality as I would by being extremely meticulous.
I'm like a bot following 30+ threads concurrently, sometimes it's fun, sometimes it feels like playing casino, sometimes it's boring, but this is truly an insane era if you have the funding for it, obviously we stack many MANY accounts in rotation 24/7, equivalent in API cost by myself is about 100K$+ (a month) but we pay only a fraction of that cost thanks to the plans.
PS: I have 8 monitors in front of me to manage all that (portable monitors stacked together).
Hope it works and you post about it.
I still haven't seen a single person who actually cares about the environment and has willingly made significant sacrifices for it, who clamors about the environmental cost of AI. Every time I see someone do it it's someone who never cared about this before, and still doesn't really. Who buys plenty of new clothes and furniture, loves a good burger, has the latest iPhone, flies 4 times per year.
Maybe you're the unicorn in which case fair enough, you've earned the right to feel disgusted.
I have no sympathy for OpenAI or Anthropic as corporations, but if these are the new tools of the trade, then platform abuse like GP is bragging about serves only to destroy the livelihoods of the rest of us who are content to use our fair share.
There’s no such thing as a free lunch, and the bill always comes at the end.
Don't be surprised if/when people ignore your AI slop
If so, I’m not sure I’d even consider that a problem. If the goal is for it to give you a query to run, and you ask it “let’s do it in a transaction”, it’s a reasonable thing for it to simply inform you, “yeah you can just type begin first” since it’s assuming you’re going to be pasting the query in anyway. And yeah, it does use fewer tokens, assuming the query was long. Similar to how, if it gave me a command to run, and I say “I’m getting a permission denied”, it would be reasonable for it to say “yeah do it as root, put sudo before the command”, and it’s IMO reasonable if it didn’t repeat the whole thing verbatim just with the word “sudo” first.
But if the context was that you actually expected it to run the query for you, and instead it just said “here, you run it”, then yeah that’s lazy and I’d understand the shock.
All earnesty aside, OpenAI’s oddly specific singular focus on “intelligence per token” (also in the benchmarks) that literally noone else pushes so hard eerily reminds me of Apple’s Macbook anorexia era pre-M1. One metric to chase at the cost of literally anything else. GPT-5.3+ are some of the smartest models out there and could be a pleasure to work with, if they weren’t lazy bastards to the point of being completely infuriating.
I know it's only on a single benchmark, but I dont understand how it can be so bad...
The models not availble on copilot were tested through opencode (max reasoning) and deepseek v4 was tested through Cline (with max reasoning too).
I really like this benchmarking. Have you evaluated the judge benchmark somehow? I'd love to setup my own similar benchmark.
I haven't evaluated the judge benchmark. You have everything needed in the repo to do so though, so be my guest. It took me a bit of time to put all this together and won't have much more time to dedicate to it before a couple of weeks.
BTW, if you explore the repo, sorry for all the French files...
Your prompt is extremely slim yet you score it on a bunch of features.
The eval prompt is quite extensive: https://github.com/guilamu/llms-wordpress-plugin-benchmark/b...
I personally develop with very detailed spec, and I don’t want nothing more and nothing less compared to the spec.
I found 5.4/5.5 much better at following spec while Opus makes some things up, which aligns with your benchmark but that makes 5.4/5.5 better for me while worse for you.
What strike me as very strange though is that 0 model were able to just use the search input already present in GravitYForms forms list page and all created a second input.
Also, I know it's not in the prompt, but adding a ctrl+f shortcut to a search input? Is that that crazy? I don't know.
Input: $5/M tokens at <=272K, $10/M tokens above 272K.
Output: $30/M tokens at <=272K, $45/M tokens above 272K.
Cache read: $0.50/M tokens at <=272K, $1/M tokens above 272K.
Significantly more expensive than Opus 4.7 beyond 272K and at least in my tasks, I haven't seen the model that much more token efficient, certainly not to such a degree that it'd compensate this difference. GPT-5.4 had a solid context window at 400k with reliable compaction, both appear somewhat regressed, though still to early to truly say whether compaction is less reliable. Also, I have found frontend output to still skew towards that one very distinct, easily noticeable, card laden, bluesy hue overindulged template that made me skeptical of Horizon Alpha/Beta pre GPT-5s release. Ended up doing amazing at the time for task adherence, which made it very useful for me outside that one major deficit. The fact that GPT-5.5 is still so restricted in that area is weird considering it's supposed to be an entirely new foundation.
>API deployments require different safeguards and we are working closely with partners and customers on the safety and security requirements for serving it at scale.
And now this. I guess one day counts as "very soon." But I wonder what that meant for these safeguards and security requirements.
> In 2023, the company was preparing to release its GPT-4 Turbo model. As Sutskever details in the memos, Altman apparently told Murati that the model didn’t need safety approval, citing the company’s general counsel, Jason Kwon. But when she asked Kwon, over Slack, he replied, “ugh . . . confused where sam got that impression.”
Lots of cases where Altman hass not been entirely forthcoming about how important (or not) safety is for OpenAI. https://www.newyorker.com/magazine/2026/04/13/sam-altman-may... (https://archive.is/a2vqW)
There is too much and there are too many, and some of their takes don’t fly if you use Claude daily.
If this sounds crazy to you, though, I'd like to know, and understand why. I miss ChatGPT/Codex.
Cheaper and slower than Opus.
Live decision and heavier agentic evals will continue being uploaded for 24 hours but I don't expect its leaderboard position to change at this point.
GPT 5.5 is the most intelligent public model. And significantly faster than its predecessor.
Just ask it about an event that happened shortly before Dec 1, 2025. Sporting event, preferably.
could be they do it intentionally to encourage more tool calls/searches or for tuning reasons
The proper way to figure out the real cutoff date is to ask the model about things that did not exist or did not happen before the date in question.
A few quick tests suggest 5.5's general knowledge cutoff is still around early 2025.
Whatever it is, the cutoff date reporting discrepancy isn't new. Back when Musk was making headlines about buying/not buying Twitter, I was able to find recent-ish related news that was published well after the bot's stated cutoff date.
ChatGPT was not yet browsing/searching/using the web at that point. That tool didn't come for another year or so.
Easiest Turing test ever...
A better test is something like "what is the latest version of NumPy?"
You're probably better off asking something like "what are the most notable changes in version X of NumPy?" and repeating until you find the version at which it says "I don't know" or hallucinates.
They all roughly produce junior developer-level code, continue to have mental breakdowns in their “thinking” stage, occasionally hallucinate things, delete pieces of code/docs they don’t understand or don’t like, use 1.5 times the necessary words to explain things when generating docs and so on.
I'm now testing "avoid sycophancy, keep details short and focus on the facts" in my AGENTS.md files.
AI is like having the greatest developer who ever lived, but she is always on 4 beers.
All the AI players definitely seem to be trying to claw more money out of their users at the moment.
(Note, that stops being true at higher reasoning levels, where our observed total cost goes up ~2-3x.)
[1] https://x.com/voratiq/status/2047737190323769488?s=20
30/180 usd on Openrouter. Did I miss something?
4.6 did very well. 90% perfect on first try, got to 100% with just a few followups. 4.7 failed horribly. First produced garbage output and claimed it was done, admitted it did that when called out, proceeded to work at it a lot longer and then IT GAVE UP. GPT 5.5 codex was shockingly good. Achieved 90% perfect on first try in about a fourth of the time. Got to 100% faster and with fewer follow-ups.
I’m impressed.
Would be interesting if you ran your same test with Deepseek v4 and some of the other Chinese models.
Gave it two very long-running problems I haven't had the courage to work on in the last 2.5 years, solved each within an hour.
- An incremental streaming JSON decoder that can optionally take a list of keys to stop decoding after. 1800 LOC about 30 minutes later, and now my local-first apps first sync time is 0.8s instead of 75s when there's 1.5 GB of data locally.
- Flutter Web can compile to WASM and then render via Skia WASM. I've been getting odd crashes during rapid animation for months. In an hour, it got Skia WASM checked out, building locally, a Flutter test script, and root caused the issue to text shadows and font glyphs (technically, not solved yet, I want to get to the point we have Skia / Flutter patch(es))
If you told me a week ago that an LLM could do either of these, without heavy guidance, I'd be stunned. And I regularly push them to limits, ex. one of Opus' last projects was a tolerant JSON decoder, and it ended up being 8% faster than the one built-in to Dart/Flutter, which has plenty of love and attention. (we're cheating a little, that's why it's faster. TL;DR: LLMs will emit control characters in JSON and that's fine for me, treating them as fine means file edit error rates go from ~2% to 0%)
I just wish it was cheaper, but, don't we all...
Either Opus 4.7 miscounts reasoning tokens, or it's A LOT more efficient than GPT 5.5
I thought they made GPT 5.5 more token efficient than 5.4, but it uses 2x the reasoning tokens.
[0]: https://aibenchy.com/compare/openai-gpt-5-5-medium/openai-gp...
In my place for example, a lot of doctors are using ChatGPT both to search diagnosis and communicate with non-English speaking patients.
Even yourself, when you want to learn about one disease, about some real-world threats, some statistics, self-defense techniques, etc.
Otherwise it's like blocking Wikipedia for the reason that using that knowledge you can do harmful stuff or read things that may change your mind.
Freedom to read about things is good.
I think that's the problem. Who's going to claim responsibility when ChatGPT hallucinates or mistranslates a patient's diagnosis and they die? For OpenAI, this would at best be a PR nightmare, so that's why they have safeguards.
A doctor who gets refused by ChatGPT doesn't stop needing to communicate with the patient; they fall back to a worse option (Google Translate, a family member interpreting, guessing). Refusal isn't safety, it's liability-shifting dressed up as safety.
If there's no doctor, no interpreter, no pharmacist, just a person with a sick kid and a phone, then "refuse and redirect to a professional" is advice from a world that doesn't exist for them. The refusal doesn't send them to a better option; there is no better option, it's a large majority of people on this planet.
Hell is paved of good intentions, but open-education and unlimited access to knowledge is very good.
It doesn't change the human nature of some people, bad people stay bad, good people stay good.
About PR, they're optimizing for not being the named defendant in a lawsuit or the subject of a bad news cycle, it's self-interest wearing benevolence as a costume.
This is because harms from answering are punishable (bad PR, unhappy advertisers, unhappy investors, unhappy politicians / dictators, unhappy lobbies, unhappy army, etc); but harms from refusing are invisible and unpunished.
I think AI proves the contrary. There are plenty of examples of things that are getting worse because of technological advancement, particularly AI. Software quality, writing, online discourse, misinformation have all suffered over the last few years. I truly believe the internet is a worse place than it was 5 years ago, and I can't imagine bringing that to medicine would work out differently.
The medical system shouldn't rely on falling back to crappy workarounds, it should aspire to build the best system it reasonably can.
I had a choice better a doctor that used AI or not, I would much prefer one that did...
Only the liability shifts from OpenAI to them.
Furthermore, where the alternative to a licensed professional was nothing, or a random untrained person or a weak professional, then it's harming the user on the pretext of protecting him.
(like in the other mentioned contexts).
-> You are in China, you go to emergency, nobody speaks your language
Move hands ? DeepSeek is better than using hands, even Baidu Translate, ChatGPT or whatever you find.
Other solutions are theoretically nice on paper but almost delusional.
An imperfect solution is better than no solution.
==
Similarly, a deaf-person is theorically better with a certified interpreter that can talk with the hands, but they may prefer voice-recognition software or AI tools.
(or... talking with hands is more confusing and annoying or less understandable for them).
Of course ChatGPT transcription can have issues, but that's the difference between the real-world and Silicon Valley's disconnected lawyers world.
==
If ChatGPT says: "sorry I won't be able, please go to see a licensed interpreter, good luck!" then it's just OpenAI trying to save their asses, at your risk/expense.
If you have a choice, you can make the choice, and you can double-check what is said. In other cases, you have no choice, nothing to check, only problems but no hints of solutions.
This is why openness is important.
https://www.england.nhs.uk/interpreting/