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

GPT-5.4

openai.com

Post reply on HN
Print view View original

691–700 of 868 posts

Re: GPT-5.4

#691

Post by startages »

creamyhorror wrote: post #223

I've only used 5.4 for 1 prompt (edit: 3@high now) so far (reasoning: extra high, took really long), and it was to analyse my codebase and write an evaluation on a topic. But I found its writing and analysis thoughtful, precise, and surprisingly clearly written, unlike 5.3-Codex. It feels very lucid and uses human phrasing. It might be my AGENTS.md requiring clearer, simpler language, but at least 5.4's doing a good…

I thought I had something wrong within my setup, I could never use Codex 5.3 while everyone else was praising it. It uses some weird terms and complex jargon and doesn't really make it clear what it was doing or planning to do unlike Opus which makes things clear, this allows me to give accurate feedback and change plans and make proper decision.

Re: GPT-5.4

#692

Post by rrr_oh_man »

jryio wrote: post #16

1 million tokens is great until you notice the long context scores fall off a cliff past 256K and the rest is basically vibes and auto compacting.

It's the same now with Gemini as well. Unfortunately. :(

Re: GPT-5.4

#693

Post by mindplunge »

tedsanders wrote: post #59
minimaxir wrote: post #8

The marquee feature is obviously the 1M context window, compared to the ~200k other models support with maybe an extra cost for generations beyond >200k tokens. Per the pricing page, there is no additional cost for tokens beyond 200k: https://openai.com/api/pricing/ Also per pricing, GPT-5.4 ($2.50/M input, $15/M output) is much cheaper than Opus 4.6 ($5/M input, $25/M output) and Opus has a penalty for its beta >200…

Yeah, long context vs compaction is always an interesting tradeoff. More information isn't always better for LLMs, as each token adds distraction, cost, and latency. There's no single optimum for all use cases. For Codex, we're making 1M context experimentally available, but we're not making it the default experience for everyone, as from our testing we think that shorter context plus compaction works best for most p…

Frontend work with large component libraries. When I'm refactoring shared design system components, things like a token system that touches 80+ files, compaction tends to lose the thread on which downstream components have already been updated vs which still need changes. It ends up re-doing work or missing things silently.

The model holds "what has been updated" well at the start of a session. After compaction, it reconstructs from summaries, and that reconstruction is lossy exactly where precision matters most: tracking partially-complete cross-file operations.

1M context isn't about reading more, it's about not forgetting what you already did halfway through.

Re: GPT-5.4

#694

Post by swores »

apetresc wrote: post #382
glenstein wrote: post #299

Earlier quoted context omitted.

Wow, that's diametrically the opposite point: the cost is *extra*, not free.

Diametrically opposite to tokens beyond 200K being literally free? As in, you only pay for the first 200K tokens and the remaining 800K cost $0.00? I don't think that's a fair reading of the original post at all, obviously what they meant by "no cost" was "no increase in the cost".

I can see that's what they mean now that I've read the replies, but when I first read that top comment I too parsed it as meaning 201k would cost the same as 999k (which admittedly did seem strange, hence I read the replies to confirm and sure enough that's not actually the case!)

Re: GPT-5.4

#695

Post by Tarq0n »

irishcoffee wrote: post #471
creamyhorror wrote: post #456

Earlier quoted context omitted.

I'm not sure if the model (under its temperature/other settings) produces deterministic responses. But I do think models' style and phrasing are fairly changeable via AGENTS.md-style guidelines. 5.4's choice of terms and phrasing is very precise and unambiguous to me, whereas 5.3-Codex often uses jargon and less precise phrases that I have to ask further about or demand fuller explanations for via AGENTS.md.

So sharing markdown files is functionally useless, or no?

No it's just stochastic like everything about LLMs. The md file will bias results towards a certain set of outcomes.

Re: GPT-5.4

#696

Post by hdjrudni »

yieldcrv wrote: post #418
fhrow4484 wrote: post #360

Earlier quoted context omitted.

https://static0.anpoimages.com/wordpress/wp-content/uploads/...

Preview Road (only choice, and last preview was deprecated without warning)

If the last preview was 'deprecated', it's still usable. So you have two choices.

Peeve of mine when people say 'deprecated' but really they mean 'discontinued' or 'deleted'.

Things don't instantly disappear when they're deprecated.

Re: GPT-5.4

#697

Post by dahcryn »

tedsanders wrote: post #59
minimaxir wrote: post #8

The marquee feature is obviously the 1M context window, compared to the ~200k other models support with maybe an extra cost for generations beyond >200k tokens. Per the pricing page, there is no additional cost for tokens beyond 200k: https://openai.com/api/pricing/ Also per pricing, GPT-5.4 ($2.50/M input, $15/M output) is much cheaper than Opus 4.6 ($5/M input, $25/M output) and Opus has a penalty for its beta >200…

Yeah, long context vs compaction is always an interesting tradeoff. More information isn't always better for LLMs, as each token adds distraction, cost, and latency. There's no single optimum for all use cases. For Codex, we're making 1M context experimentally available, but we're not making it the default experience for everyone, as from our testing we think that shorter context plus compaction works best for most p…

I would like to counteract your statement that each token adds a distraction.

In our experiments, we see a surprising benefit to rewriting blocks to use more tokens, especially long lists etc..

E.g. compare these two options

"The following conditions are excluded from your contract - condition A - condition B ... - condition Z"

The next one works better for us:

"The following conditions are excluded from your contract - condition A is excluded - condition B is excluded ... - condition Z is excluded"

And we now have scripts to rewrite long documents like this, explicitly adding more tokens. Would you have any opinion on this?

Re: GPT-5.4

#698

Post by sagarpatil »

mattas wrote: post #5

"GPT‑5.4 interprets screenshots of a browser interface and interacts with UI elements through coordinate-based clicking to send emails and schedule a calendar event." They show an example of 5.4 clicking around in Gmail to send an email. I still think this is the wrong interface to be interacting with the internet. Why not use Gmail APIs? No need to do any screenshot interpretation or coordinate-based clicking.

Or CLI.

Re: GPT-5.4

#699

Post by AmazingTurtle »

I just tried that in Codex CLI. With /fast mode enabled. Observations:

1. Fast mode ain't that fast

2. Large context * Fast * Higher Model Base Price = 8x increase over gpt-5.3-codex

3. I burnt 33% of my 5h limit (ChatGPT Business Subscription) with a prompt that took 2 minutes to complete.

Re: GPT-5.4

#700

Post by fvv »

minimaxir wrote: post #8

The marquee feature is obviously the 1M context window, compared to the ~200k other models support with maybe an extra cost for generations beyond >200k tokens. Per the pricing page, there is no additional cost for tokens beyond 200k: https://openai.com/api/pricing/ Also per pricing, GPT-5.4 ($2.50/M input, $15/M output) is much cheaper than Opus 4.6 ($5/M input, $25/M output) and Opus has a penalty for its beta >200…

imo , the main feature is /fast ... who use 1M context and for what? the model become dumber already at 200K.. it's better to manage the context , and since 5.3, codex is very good at managing it
Post reply on HN