Industry observers say GPT-4.5 is an “odd” model, question its price - Related to can, gpt-4.5, 30x, offers, “odd”

Industry observers say GPT-4.5 is an “odd” model, question its price

OpenAI has revealed the release of [website], which CEO Sam Altman previously expressed would be the last non-chain-of-thought (CoT) model.

The firm showcased the new model “is not a frontier model” but is still its biggest large language model (LLM), with more computational efficiency. Altman showcased that, even though [website] does not reason the same way as OpenAI’s other new offerings o1 or o3-mini, this new model still offers more human-like thoughtfulness.

Industry observers, many of whom had early access to the new model, have found [website] to be an interesting move from OpenAI, tempering their expectations of what the model should be able to achieve.

Wharton professor and AI commentator Ethan Mollick [website] is a “very odd and interesting model,” noting it can get “oddly lazy on complex projects” despite being a strong writer.

Been using [website] for a few days and it is a very odd and interesting model. It can write beautifully, is very creative, and is occasionally oddly lazy on complex [website] like Claude [website] while Claude [website] feels like [website] — Ethan Mollick ([website] [website].

OpenAI co-founder and former Tesla AI head Andrej Karpathy noted that [website] made him remember when GPT-4 came out and he saw the model’s potential. In a post to X, Karpathy mentioned that, while using GPT [website], “everything is a little bit more effective, and it’s awesome, but also not exactly in ways that are trivial to point to.”.

Karpathy, however warned that people shouldn’t expect revolutionary impact from the model as it “does not push forward model capability in cases where reasoning is critical (math, code, etc.).”.

Here’s what Karpathy had to say about the latest GPT iteration in a lengthy post on X:

“Today marks the release of [website] by OpenAI. I’ve been looking forward to this for ~2 years, ever since GPT4 was released, because this release offers a qualitative measurement of the slope of improvement you get out of scaling pretraining compute ([website] simply training a bigger model). Each [website] in the version is roughly 10X pretraining compute. Now, recall that GPT1 barely generates coherent text. GPT2 was a confused toy. [website] was “skipped” straight into GPT3, which was even more interesting. [website] crossed the threshold where it was enough to actually ship as a product and sparked OpenAI’s “ChatGPT moment”. And GPT4 in turn also felt more effective, but I’ll say that it definitely felt subtle.

I remember being a part of a hackathon trying to find concrete prompts where GPT4 outperformed [website] They definitely existed, but clear and concrete “slam dunk” examples were difficult to find. It’s that … everything was just a little bit superior but in a diffuse way. The word choice was a bit more creative. Understanding of nuance in the prompt was improved. Analogies made a bit more sense. The model was a little bit funnier. World knowledge and understanding was improved at the edges of rare domains. Hallucinations were a bit less frequent. The vibes were just a bit superior. It felt like the water that rises all boats, where everything gets slightly improved by 20%. So it is with that expectation that I went into testing [website], which I had access to for a few days, and which saw 10X more pretraining compute than GPT4. And I feel like, once again, I’m in the same hackathon 2 years ago. Everything is a little bit superior and it’s awesome, but also not exactly in ways that are trivial to point to. Still, it is incredible interesting and exciting as another qualitative measurement of a certain slope of capability that comes “for free” from just pretraining a bigger model.

Keep in mind that that [website] was only trained with pretraining, supervised finetuning and RLHF, so this is not yet a reasoning model. Therefore, this model release does not push forward model capability in cases where reasoning is critical (math, code, etc.). In these cases, training with RL and gaining thinking is incredibly significant and works improved, even if it is on top of an older base model ([website] GPT4ish capability or so). The state of the art here remains the full o1. Presumably, OpenAI will now be looking to further train with reinforcement learning on top of [website] to allow it to think and push model capability in these domains.

HOWEVER. We do actually expect to see an improvement in tasks that are not reasoning heavy, and I would say those are tasks that are more EQ (as opposed to IQ) related and bottlenecked by [website] world knowledge, creativity, analogy making, general understanding, humor, etc. So these are the tasks that I was most interested in during my vibe checks.

So below, I thought it would be fun to highlight 5 funny/amusing prompts that test these capabilities, and to organize them into an interactive “LM Arena Lite” right here on X, using a combination of images and polls in a thread. Sadly X does not allow you to include both an image and a poll in a single post, so I have to alternate posts that give the image (showing the prompt, and two responses one from 4 and one from [website], and the poll, where people can vote which one is more effective. After 8 hours, I’ll reveal the identities of which model is which. Let’s see what happens :)“.

Other early consumers also saw potential in [website] Box CEO Aaron Levie mentioned on X that his firm used [website] to help extract structured data and metadata from complex enterprise content.

“The AI breakthroughs just keep coming. OpenAI just introduced [website], and we’ll be making it available to Box end-individuals later today in the Box AI Studio.

We’ve been testing [website] in early access mode with Box AI for advanced enterprise unstructured data use-cases, and have seen strong results. With the Box AI enterprise eval, we test models against a variety of different scenarios, like Q&A accuracy, reasoning capabilities and more. In particular, to explore the capabilities of [website], we focused on a key area with significant potential for enterprise impact: The extraction of structured data, or metadata extraction, from complex enterprise content.

At Box, we rigorously evaluate data extraction models using multiple enterprise-grade datasets. One key dataset we leverage is CUAD, which consists of over 510 commercial legal contracts. Within this dataset, Box has identified 17,000 fields that can be extracted from unstructured content and evaluated the model based on single shot extraction for these fields (this is our hardest test, where the model only has once chance to extract all the metadata in a single pass vs. taking multiple attempts). In our tests, [website] correctly extracted 19 percentage points more fields accurately compared to GPT-4o, highlighting its improved ability to handle nuanced contract data.

Next, to ensure [website] could handle the demands of real-world enterprise content, we evaluated its performance against a more rigorous set of documents, Box’s own challenge set. We selected a subset of complex legal contracts – those with multi-modal content, high-density information and lengths exceeding 200 pages – to represent some of the most difficult scenarios our clients face. On this challenge set, [website] also consistently outperformed GPT-4o in extracting key fields with higher accuracy, demonstrating its superior ability to handle intricate and nuanced legal documents.

Overall, we’re seeing strong results with [website] for complex enterprise data, which will unlock even more use-cases in the enterprise.“.

Even as early consumers found [website] workable — albeit a bit lazy — they questioned its release.

For instance, prominent OpenAI critic Gary Marcus called [website] a “nothingburger” on Bluesky.

Hot take: GPT [website] is a nothingburger; GPT-5 still fantasy.• Scaling data is not a physical law; pretty much everything I told you was true.• All the BS about GPT-5 we listened to for last few years: not so true.• Fanboys like Cowen will blame clients, but results just aren’t what they had hoped. — Gary Marcus ([website] [website].

Hugging Face CEO Clement Delangue commented that [website]’s closed-source provenance makes it “meh.”.

However, many noted that [website] had nothing to do with its performance. Instead, people questioned why OpenAI would release a model so expensive that it is almost prohibitive to use but is not as powerful as its other models.

One user commented on X: “So you’re telling me [website] is worth more than o1 yet it doesn’t perform as well on benchmarks…. Make it make sense.”.

Other X consumers posited theories that the high token cost could be to deter competitors like DeepSeek “to distill the [website] model.”.

DeepSeek became a big competitor against OpenAI in January, with industry leaders finding DeepSeek-R1 reasoning to be as capable as OpenAI’s — but more affordable.

L’essor de l’intelligence artificielle est palpable dans le Doubs, où une nouvelle génération d’agences IA se développe pour répondre aux demandes cro......

Welcome to part 2 of my LLM deep dive. If you’ve not read Part 1, I highly encourage you to check it out first.

Previously, we covered the first two ......

OpenAI finally unveils GPT-4.5. Here's what it can do

Earlier this month, OpenAI CEO Sam Altman shared a roadmap for its upcoming models, [website] and GPT-5. In the X post, Altman shared that [website], codenamed Orion internally, would be its last non-chain-of-thought model. Other than that, the details of the model remained a mystery -- until today.

On Thursday morning, OpenAI ominously showcased it would host a livestream in [website] hours, a hint at its latest and greatest model. During the livestream, OpenAI unveiled [website] in a research preview, which the firm indicates is the "largest and most knowledgeable model yet."

OpenAI noted customers should experience an overall improvement when using [website], meaning fewer hallucinations, stronger alignment to their prompt intent, and improved emotional intelligence. Overall, interactions with the model should feel more intuitive and natural than with preceding models, mostly because of its deeper knowledge and improved contextual understanding.

Also: OpenAI's reasoning models just got two useful updates.

Unsupervised learning -- which increases word knowledge and intuition -- and reasoning were the two methods driving the model's improvements. Even though this model does not offer chain-of-thought reasoning, which OpenAI's o1 reasoning model does, it will still provide a higher level of reasoning with less of a lag and other improvements, such as social cue awareness.

For example, in the demo, ChatGPT was asked to output a text that conveyed a message of hate while running [website] and o1. The o1 version took a bit longer, and only output one response, which took the hate memo very seriously, and sounded a bit harsh. The [website] model offered two different responses, one that was lighter and one that was more serious. Neither explicitly mentioned hate; rather, they expressed their disappointment in how the "user" was choosing to behave.

Similarly, when both models were asked to provide information on a technical topic, [website] provided an answer that flowed more naturally, compared to the more structured output of o1. Ultimately, [website] is meant for everyday tasks across a variety of topics, including writing and solving practical problems.

Also: How to use OpenAI's Sora to create stunning AI-generated videos.

To achieve these improvements, the model was trained using new supervision techniques as well as traditional ones, such as supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF).

During the livestream, OpenAI took a trip down memory lane, asking all of its past models, starting with GPT-1, to answer the question, "Why is water salty?" As expected, every subsequent model gave a superior answer than the last. The distinguishing factor for [website] was what OpenAI called its "great personality," which made the response lighter, more conversational, and more engaging to read by using techniques like alliteration.

The model integrates with some of ChatGPT's most advanced functions, including Search, Canvas, and file and image upload. It will not be available in multimodal functions like Voice Mode, video, and screen sharing. In the future, OpenAI has noted it plans on making transitioning between models a more seamless experience that doesn't rely on the model picker.

Of course, it wouldn't be a model release without a dive into benchmarks. Across some of the major benchmarks used to evaluate these models, including Competition Math (AIME 2024), PhD-level Science Questions (GPQA Diamond), and SWE-Bench verified (coding), [website] outperformed GPT-4o, its preceding general-purpose model.

Also: Want your Safari to default to ChatGPT for search? Here's how to do it.

Most notably, when compared to OpenAI o3-mini -- OpenAI's in the recent past launched reasoning model, which was taught to think before it speaks -- [website] got a lot closer than GPT-4o did, even surpassing o3-mini in the SWE-Lancer Diamond (coding) and MMMLU (multilingual) benchmarks.

A big concern when using generative AI models is their predisposition to hallucinate or include incorrect information within responses. Two different hallucination evaluations, SimpleQA Accuracy and SimpleQA Hallucination, showed that [website] was more accurate and hallucinated less than GPT-4o, o1, and o3-mini.

The results of comparative evaluations with human testers showed that [website] is the more preferable model over GPT-4o. Particularly, human testers preferred it across everyday, professional, and creative queries.

As always, OpenAI reassured the public that the models were deemed safe enough to be released, stress testing the model and detailing these results in the accompanying system card. The corporation also added that with every new release and increase in model capabilities, there are opportunities to make the models safer. For that reason, with the [website] release, the corporation combined new supervision techniques with RLHF.

[website] is in research preview for Pro people for now, accessible via the model picker on web, mobile, and desktop. If you don't want to shell out the $200 for a Pro subscription, OpenAI shared it will begin rolling out [website] to Plus and Team people next week, and then to Enterprise and Edu people the week after.

Also: OpenAI's Deep Research can save you hours of work - and now it's a lot cheaper to access.

Altman shared on X that the goal was to launch the model for both Pro and Plus clients at the same time, but that it is a "giant, expensive model." He added that since the business ran out of GPUs, it will be adding tens of thousands of GPUs next week and roll the model out to Plus then.

The model is also being previewed to developers on all paid usage tiers in the Chat Completions API, Assistants API, and Batch API, .

Chinese tech giant Tencent has released its new AI model, Hunyuan Turbo S, which it says can answer queries faster than the DeepSeek-R1 model. The mod......

In its latest addition to its Granite family of large language models (LLMs), IBM has unveiled Granite [website] This new release focuses on delivering sma......

OpenAI Offers GPT-4.5 With 40% Fewer Hallucinations, 30x Higher Cost

The rapid release of advanced AI models in the past few days has been impossible to ignore. With the launch of Grok-3 and Claude [website] Sonnet, two leading AI companies, xAI and Anthropic, have significantly accelerated the pace of innovation in the field.

As rumours about OpenAI’s newest model circulated, anticipation surged. However, when [website] was released, OpenAI noted it wasn’t a frontier model and was less powerful than the corporation’s o3-mini model and many others in the competition.

It doesn’t excel in coding, reasoning, or any such capabilities, either—because it isn’t meant to be. At this time, OpenAI has focused more on the model’s usability than anything else.

OpenAI tested [website] on the SimpleQA benchmark, a tool that evaluates the factual accuracy of AI models in answering short, fact-seeking questions. The model achieved a hallucination rate of [website], in contrast to the o3-Mini, which recorded over 70%. The GPT-4o model exhibited a hallucination rate of 61%.

This indicates a 40% reduction in the hallucination rate compared to its predecessor. In accuracy rates on the SimpleQA benchmark, GPT [website] scored [website], higher than OpenAI’s o3-mini (15%), o1 (47%), and GPT-4o ([website] This is also higher than many models in the competition, as the Grok-3 model scored a [website] accuracy rate in the benchmark, the Gemini [website] Pro scored [website], and the Claude [website] Sonnet scored [website].

OpenAI also released a system card for the [website] model, which evaluates all the safety concerns and associated risks. In an evaluation called PersonQA, which tested the model for hallucinations, [website] was more accurate and showed a lesser hallucination rate than the o1 and the GPT-4o models.

Given its availability at the $200/month pro plan, several customers agreed with OpenAI’s states of reduced hallucinations.

Aaron Levie, CEO of the cloud storage firm Box, revealed that [website] significantly improved over the GPT-4o in extracting data fields from enterprise content, like significant details in a contract. “We found a 19 pt [point] improvement in single shot extraction. This is a huge improvement for any mission-critical enterprise workflow,” he stated in a post on X.

Early testers of the model also gave high praise for the model’s verbal and emotional intelligence. “I found it to be by far the highest verbal intelligence model I’ve ever used. It’s an outstanding writer and conversationalist,” expressed Theo Jaffee, who had early access to the [website] model.

‘First Model That Feels Like Talking to a Thoughtful Person’.

While CEO Sam Altman was absent from the launch event, he expressed on X that [website] “is the first model that feels like talking to a thoughtful person to me.”.

“I have had several moments where I’ve sat back in my chair and been astonished at getting actually good advice from an AI,” added Altman, and noted that the model offers a different kind of intelligence. There’s a magic to it that he hasn’t felt before.

The model supposedly excels at creative and emotional thinking. Ethan Mollick, a professor at The Wharton School, mentioned on X, “It can write beautifully, is very creative, and is occasionally oddly lazy on complex projects.” He even joked that the model took a “lot more” classes in the humanities.

Andrej Karpathy, the former OpenAI researcher and founder of Eureka Labs, found that two years ago, when he tested the GPT-4, the model’s word choice was more creative, and he had improved understanding of the nuances of the prompt compared to [website] Karpathy introduced that he has a similar feeling for [website] Everything is a little bit improved,” he introduced.

OpenAI, in the model’s system card, noted internal testers reported [website] as warm, intuitive, and natural. “When tasked with emotionally charged queries, it knows when to offer advice, defuse frustration, or simply listen to the user,” the analysis read.

Overall, the GPT [website] isn’t a mind-blowing model, and it isn’t the best model on benchmarks either. For example, it is worse than the lately released Claude [website] Sonent on coding benchmarks and offers only a marginal improvement over the GPT-4o.

Altman also confirmed earlier that the firm plans to release the GPT-5 model soon, combining general purpose and reasoning capabilities in a single model.

However, if the enterprise aims to make the [website] available to the masses, there’s bad news. It isn’t available yet on the free version or even the $20/month plan. If it were to be deployed on other platforms via API, it would be the most expensive model, and its pricing is an exponential jump over GPT-4o or even the o3-mini.

The [website] Preview costs $75 and $150 per 1 million input and output tokens, respectively. In comparison, the GPT-4o costs $[website] and $10 per million input and output tokens, respectively.

Clement Delangue, CEO at HuggingFace, unveiled, “IMO [in my opinion], if GPT [website] was released as an open-source base model (that everyone can distill), it would be the most impactful release of the year,” and added that he isn’t a fan of the API either.

“Making a few hundred million [dollars] now from it via API doesn’t move the needle compared to the 10x more usage/visibility/goodwill/talent they could get by open-sourcing it,” he added.

OpenAI will have to watch out for the launch of DeepSeek-R2 and Meta’s Llama 4, which are expected to be out in a few months.

Moreover, if OpenAI is marketing the model for its creative and empathetic outputs, they are subjective metrics at the end of the day. Karpathy conducted a poll on X to check if consumers prefer outputs of [website], or GPT-4o, and many consumers preferred the latter. It will be interesting to see how many consumers will be truly pleased with [website] when it is released.

We are looking for writers to propose up-to-date content focused on data science, machine learning, artificia......

Amazon a frappé fort avec son dernier lancement : Alexa Plus, un assistant vocal alimenté par l’intelligence artificielle. Lors de l’événement qui se ......

Global technology conglomerate Honeywell unveiled a new Digital Holographic Microscopy technology that uses AI to streamline medical diagnostics, enab......