Live data from Hacker News

GPT-4.5

openai.com

111–120 of 1001 posts

Re: GPT-4.5

#111
post #81
post #24

Considering both this blog post and the livestream demos, I am underwhelmed. Having just finished the stream, I had a real "was that all" moment, which on one hand shows how spoiled I've gotten by new models impressing me, but on another feels like OpenAI really struggles to stay ahead of their competitors. What has been shown feels like it could be achieved using a custom system prompt on older versions of OpenAIs m…

> How could they justify that asking price? They're still selling $1 for <$1. Like personal food delivery before it, consumers will eventually need to wake up to this fact - these things will get expensive, fast.

I generally question how wide spread willingness to pay for the most expensive product is. And will most users of those who actually want AI go with ad ridden lesser models...

Re: GPT-4.5

#112
Can it be self-hosted? Many institutions and organizations are hesitant to use AI because concerns of data leaking over chatbot. Open models, on the other hand, can be self-hosted. There is a deepseek arm race in other part of the world. Universities are racing to host their own deepseek. Hospitals, large businesses, local governments, even courts are deploying or showing interest in self-hosting deepseek.

Re: GPT-4.5

#113
post #26

GPT 4.5 pricing is insane: Price Input: $75.00 / 1M tokens Cached input: $37.50 / 1M tokens Output: $150.00 / 1M tokens GPT 4o pricing for comparison: Price Input: $2.50 / 1M tokens Cached input: $1.25 / 1M tokens Output: $10.00 / 1M tokens It sounds like it's so expensive and the difference in usefulness is so lacking(?) they're not even gonna keep serving it in the API for long: > GPT‑4.5 is a very large and comput…

> GPT 4.5 pricing is insane:

> I'm still gonna give it a go, though.

Seems like the pricing is pretty rational then?

Re: GPT-4.5

#115
Seeing OpenAI and Anthropic go different routes here is interesting. It is worth moving past the initial knee jerk reaction of this model being unimpressive and some of the comments about "they spent a massive amount of money and had to ship something for it..."

* Anthropic appears to be making a bet that a single paradigm (reasoning) can create a model which is excellent for all use cases.

* OpenAI seems to be betting that you'll need an ensemble of models with different capabilities, working as a single system, to jump beyond what the reasoning models today can do.

Based on all of the comments from OpenAI, GPT 4.5 is absolutely massive, and with that size comes the ability to store far more factual data. The scores in ability oriented things - like coding - don't show the kind of gains you get from reasoning models but the fact based test, SimpleQA, shows a pretty large jump and a dramatic reduction in hallucinations. You can imagine a scenario where GPT4.5 is coordinating multiple, smaller, reasoning agents and using its factual accuracy to enhance their reasoning, kind of like ruminating on an idea "feels" like a different process than having a chat with someone.

I'm really curious if they're actually combining two things right now that could be split as well, EQ/communications, and factual knowledge storage. This could all be a bust, but it is an interesting difference in approaches none-the-less, and worth considering that OpenAI could be right.

Re: GPT-4.5

#116

That presentation was super underwhelming. We got to watch them compare… the vibes? … of 4.5 vs o1. No wonder Sam wasn’t part of the presentation.

And to top it off, it costs $75.00 per 1M vibes.

Re: GPT-4.5

#117
Question: explain thread first macros in clojure in a short paragraph

GPT 4.5:

In Clojure, the *thread-first macro* (`->`) is used to improve code readability by threading a value through a series of function calls. It takes an initial value and passes it as the first argument to the next function, then continues threading the result through subsequent functions. This eliminates deeply nested expressions and makes the code more readable.

For example,

```clojure (-> 5 inc (* 2) str) ;; Equivalent to (str (* 2 (inc 5))) → "12" ```

Here, `5` is incremented to `6`, then multiplied by `2` to get `12`, and finally converted to a string.

---

Deepseek v3:

In Clojure, the `->>` (thread-last) and `->` (thread-first) macros are used to improve code readability by allowing a more linear, pipeline-like flow of operations. The `->` (thread-first) macro takes an initial value and threads it as the first argument through a series of forms. For example, `(-> x (f) (g))` translates to `(g (f x))`. This is particularly useful when working with nested function calls, as it flattens the structure and makes the sequence of transformations clearer. It’s commonly used with functions that take the threaded value as their first argument, such as collection operations or Java interop methods.

Re: GPT-4.5

#118
post #26

GPT 4.5 pricing is insane: Price Input: $75.00 / 1M tokens Cached input: $37.50 / 1M tokens Output: $150.00 / 1M tokens GPT 4o pricing for comparison: Price Input: $2.50 / 1M tokens Cached input: $1.25 / 1M tokens Output: $10.00 / 1M tokens It sounds like it's so expensive and the difference in usefulness is so lacking(?) they're not even gonna keep serving it in the API for long: > GPT‑4.5 is a very large and comput…

Input price difference: 4.5 is 30x more Output price difference:4.5 is 15x more In their model evaluation scores in the appendix, 4.5 is, on average, 26% better. I don't understand the value here.

If you ran the same query set 30x or 15x on the cheaper model (and compensated for all the extra tokens the reasoning model uses), would you be able to realize the same 26% quality gain in a machine-adjudicatible kind of way?

Re: GPT-4.5

#119
post #90

Earlier quoted context omitted.

It isn't even vaguely a distill of o1. The reasoning models are, from what we can tell, relatively small. This model is massive and they probably scaled the parameter count to improve factual knowledge retention. They also mentioned developing some new techniques for training small models and then incorporating those into the larger model (probably to help scale across datacenters), so I wonder if they are doing a bi…

You can 'distill' with data from a smaller, better model into a larger, shittier one. It doesn't matter. This is what they said they did on the livestream.

I have distilled models before, I know how it works. They may have used o1 or o3 to create some of the synthetic data for this one, but they clearly did not try and create any self-reflective reasoning in this model whatsoever.

Re: GPT-4.5

#120
post #66

Finally a scaling wall? This is apparently (based on pricing) using about an order of magnitude more compute, and is only maybe 10% more intelligent. Ideally DeepSeeks optimizations help bring the costs way down, but do any AI researchers want to comment on if this changes the overall shape of the scaling curve?

Seems on par with the existing scaling curve. If I had to speculate, this model would have been an internal-only model, but they're releasing it for PR. An optimized version with 99% of the performance for 1/10th the cost will come out later.
Post reply on HN