Useful information

Prime News delivers timely, accurate news and insights on global events, politics, business, and technology

OpenAI reveals GPT-4.5 in the midst of the burst of new launches of AI models

Unlock the editor’s summary for free

Operai has launched its largest artificial intelligence model with less “hallucinations”, the last one in a large number of AI launches by US technological groups.

The company based in San Francisco presented GPT-4.5 on Thursday, its long-awaited technology update that supports its popular product chatgpt. In the first tests, their hallucination rate, where IA systems generate inaccurate information, was 37 percent compared to almost 60 percent in their GPT-4O predecessor.

With GPT-4.5, Operai continues to bet on large and large-language models despite the arrival of highly capable smaller products, such as the New Deepseek’s R1 company, which are open, cheaper and more accessible to developers.

It occurs when competition is growing within the rapid development industry, since technological groups have rushed to launch their last models in recent weeks. Anthrope revealed his sonnet Claude 3.7 on Monday, which followed the launch of Grok 3 last week, Elon Musk’s latest XAI model.

Operai in a blog post on Thursday said that GPT-4.5 had “broader knowledge and a deeper understanding in the world, which led to reduced hallucinations and greater reliability in a wide range of topics.”

“With each new order of calculation comes new capabilities,” said the company, adding that GPT-4.5 was “on the border of what is possible in not supervised learning.”

Openai has been at the forefront of a global career to lead the AI ​​industry, which increases tens of billions of investors to finance larger models with greater capabilities that require large amounts of computer power.

It is in conversations with SoftBank and other investors to raise up to $ 40 billion at an assessment of $ 300 billion, including the new money. Anthrope is also raising funds around $ 3.5 billion at an assessment of $ 60 billion, two people said with knowledge of that process.

However, the enormous costs of executing larger models have led Openai to consider withdrawing the developer’s access to GPT-4.5 in the middle of a fierce competition for rivals prices.

Openai said that while a preview of GPT-4.5 will be available for developers who pay to use OpenAI models through their application programming interface (API), this access could be revoked in the future.

The AI ​​groups generate income largely through access to the paid API and individual subscriptions. Operai said you will see how developers use the powerful model and if it is worth considering the high cost of executing it.

The company said: “GPT-4.5 is a very large and intensive model in computation, which makes it more expensive than and not a replacement for GPT-4O (its predecessor). Because of this, we are evaluating whether to continue serving in the API in the long term, since we balance the current support capabilities with the creation of future models. “

Operai’s executive president Sam Altman said that GPT-4 costs more than $ 100mn to train, and it is widely expected that such costs increase as the size and capabilities of the models are scale and require more computer power to train and execute.

In an X publication after the announcement, Altman said the company was “out of GPU”, the chips required to execute and train AI systems.

“This is not how we want to operate, but it is difficult to predict perfectly the waves of growth that lead to GPU’s scarcity,” he said, added that they expected to get more in the coming weeks.

He also stressed that GPT-4.5 did not focus on reasoning and would not exceed the industry’s reference points, but it had “a different type of intelligence and there is a magic that I have not felt before.”

Additional reports by George Hamond in San Francisco and Melissa Heikilä in London

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *