Live data from Hacker News

GPT-5

openai.com

401–410 of 1001 posts

Re: GPT-5

#401

GPT-5 knowledge cutoff: Sep 30, 2024 (10 months before release). Compare that to Gemini 2.5 Pro knowledge cutoff: Jan 2025 (3 months before release) Claude Opus 4.1: knowledge cutoff: Mar 2025 (4 months before release) https://platform.openai.com/docs/models/compare https://deepmind.google/models/gemini/pro/ https://docs.anthropic.com/en/docs/about-claude/models/overv...

with web search, is knowledge cutoff really relevant anymore? Or is this more of a comment on how long it took them to do post-training?

I wonder if it would even be helpful because they avoid the increasing AI content

Re: GPT-5

#402
I would love to see how this performs on ARC-AGI 2, zero-shot, private eval. I hope we get an update from Chollet and team regarding performance.

Re: GPT-5

#403

Going by the system card at: https://openai.com/index/gpt-5-system-card/ > GPT‑5 is a unified system . . . OK > . . . with a smart and fast model that answers most questions, a deeper reasoning model for harder problems, and a real-time router that quickly decides which model to use based on conversation type, complexity, tool needs, and explicit intent (for example, if you say “think hard about this” in the prompt).…

> While GPT‑5 in ChatGPT is a system of reasoning, non-reasoning, and router models, GPT‑5 in the API platform is the reasoning model that powers maximum performance in ChatGPT. Notably, GPT‑5 with minimal reasoning is a different model than the non-reasoning model in ChatGPT, and is better tuned for developers. The non-reasoning model used in ChatGPT is available as gpt-5-chat-latest.

https://openai.com/index/introducing-gpt-5-for-developers/

Re: GPT-5

#404
post #137

The marketing copy and the current livestream appear tautological: "it's better because it's better." Not much explanation yet why GPT-5 warrants a major version bump. As usual, the model (and potentially OpenAI as a whole) will depend on output vibe checks.

We’re at the audiophile stage of LLMs where people are talking about the improved soundstage, tonality, reduced sibilance etc

You need to burn-in your LLM by using for 100 hours before you see the true performance of it.

Re: GPT-5

#405

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

The idea is that with AGI it will then be able to self improve orders of magnitude faster than it would if relying on humans for making the advances. It tracks that the improvements are all relatively similar at this point since they're all human-reliant.

Re: GPT-5

#406

GPT-5 knowledge cutoff: Sep 30, 2024 (10 months before release). Compare that to Gemini 2.5 Pro knowledge cutoff: Jan 2025 (3 months before release) Claude Opus 4.1: knowledge cutoff: Mar 2025 (4 months before release) https://platform.openai.com/docs/models/compare https://deepmind.google/models/gemini/pro/ https://docs.anthropic.com/en/docs/about-claude/models/overv...

with web search, is knowledge cutoff really relevant anymore? Or is this more of a comment on how long it took them to do post-training?

Still relevant, as it means that a coding agent is more likely to get things right without searching. That saves time, money, and improves accuracy of results.

Re: GPT-5

#407

Did they just say they're deprecating all of OpenAI's non-GPT-5 models?

Yeah I was wondering if they meant deprecating on the ChatGPT side, but maintaining the models on their API platform, or deprecating on both.

Re: GPT-5

#408
post #315

So models are getting pretty good at oneshotting many small project ideas I've had. What's a good place to host stuff like that? Like a modern equivalent of Heroku? I used to use a VPS for everything but I'm looking for a managed solution. I heard replit is good here with full vertical integration, but I haven't tried it in years.

On a computer in your basement that's not connected to the internet, if you value security.

Re: GPT-5

#409

ChatGPT5 in this demo: > For an airplane wing (airfoil), the top surface is curved and the bottom is flatter. When the wing moves forward: > * Air over the top has to travel farther in the same amount of time -> it moves faster -> pressure on the top decreases. > * Air underneath moves slower -> pressure underneath is higher > * The presure difference creates an upward force - lift Isn't that explanation of why wings…

Its a particular type of mistake that is really interesting and telling. It is a misconception - and a common socially disseminated simplifcation. In students, these don't come from a lack of knowledge but rather from places where knowledge is structured incorrectly. Often because the phenomenon are difficult to observe or mislead when observed. Another example is heat and temperature. Heat is not temperature, but it is easy to observe them always being the same in your day to day life and so you bring that belief into a college thermodynamics course where you are learning that heat and temperature are different for the first time. It is a commonsense observation of the world that is only incorrect in technical circles

These are places where common lay discussions use language in ways that is wrong, or makes simplifcations that are reasonable but technically incorrect. They are especially common when something is so 'obvious' that experts don't explain it, the most frequent version of the concepts being explained

These, in my testing, show up a lot in LLMs - technical things are wrong when the most language of the most common explanations simplifies or obfuscates the precise truth. Often, it pretty much matches the level of knowledge of a college freshman/sophmore or slightly below, which is sort of the level of discussion of more technical topics on the internet.

Re: GPT-5

#410
Absolutely nothing new or groundbreaking. It's just a more tuned version of a basic LLM architecture.
Post reply on HN