Connect with us

Published

on

Sam Altman, CEO of OpenAI, walks from lunch during the Allen & Company Sun Valley Conference on July 6, 2022, in Sun Valley, Idaho.

Kevin Dietsch | Getty Images News | Getty Images

OpenAI announced the latest version of its primary large language model, GPT-4, on Tuesday, that it says exhibits “human-level performance” on many professional tests.

ChatGPT-4 is “larger” than previous versions, which means it has been trained on more data and has more weights in its model file, making it more expensive to run as well.

Currently, many researchers in the field believe many of the recent advancements in AI come from running ever-larger models on thousands of supercomputers in training processes that can cost tens of millions of dollars. GPT-4 is an example of an approach centering around “scaling up” to achieve better results.

OpenAI said it used Microsoft Azure to train the model; Microsoft has invested billions in the startup. OpenAI did not publish details about the specific model size or the hardware it used to train it, which could be used to recreate the model, citing “the competitive landscape.”

OpenAI’s GPT large language model powers many of the artificial intelligence demos that have been wowing people in the technology industry in the past six months, including Bing’s AI chat and ChatGPT, and the latest version is a preview of new advancements that could start filtering down to consumer products like chatbots in the coming weeks. Bing’s AI chatbot uses GPT-4, Microsoft said on Tuesday.

OpenAI says the new model will produce fewer factually incorrect answers, go off the rails and chat about forbidden topics less often, and even perform better than humans on many standardized tests.

GPT-4 performed at the 90th percentile on a simulated bar exam, the 93rd percentile on an SAT reading exam, and the 89th percentile on the SAT Math exam, OpenAI claimed.

However, OpenAI warns that the new software isn’t perfect yet and that it is less capable than humans in many scenarios. It still has a major problem with “hallucination,” or making stuff up, and isn’t factually reliable, the company said. It is still prone to insisting it is correct when it is wrong.

“GPT-4 still has many known limitations that we are working to address, such as social biases, hallucinations, and adversarial prompts,” the company said in a blog post.

“In a casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle. The difference comes out when the complexity of the task reaches a sufficient threshold—GPT-4 is more reliable, creative, and able to handle much more nuanced instructions than GPT-3.5,” OpenAI wrote in a blog post.

The new model will be available to paid ChatGPT subscribers and will also be available as part of an API which allows programmers to integrate the AI into their apps. OpenAI will charge about 3 cents for about 750 words of prompts and 6 cents for about 750 words in response.

Nvidia expanded from gaming into A.I. Now the big bet is paying off as its chips power ChatGPT

Continue Reading

Technology

Circle shares fall after stablecoin issuer says it will offer 10 million shares

Published

on

By

Circle shares fall after stablecoin issuer says it will offer 10 million shares

Circle Internet Group Initial Public Offering at the New York Stock Exchange in New York City, U.S., June 5, 2025.

NYSE

Circle Internet Group stock tumbled more than 5% in extended trading Tuesday after it said it would offer 10 million Class A shares to the public.

Of the total stock being offered, 2 million shares will be offered by Circle. The remaining 8 million shares will be sold by stockholders.

The stablecoin issuer’s shares have soared more than 450% since it went public on June 5.

As part of the offering, Circle is offering its underwriters a 30-day option to buy an additional 1.5 million shares.

Circle shares closed Tuesday up 1.3% after the company reporting its first quarterly results as a publicly traded company. While charges tied to its IPO weighed on its second-quarter results and led to a loss of $4.48 per share, it saw revenue rise 53% on the back of strong stablecoin growth.

Don’t miss these cryptocurrency insights from CNBC Pro:

Continue Reading

Technology

CoreWeave shares drop even as revenue tops estimates

Published

on

By

CoreWeave shares drop even as revenue tops estimates

Mike Intrator, co-founder and CEO of CoreWeave, speaks at the Nasdaq headquarters in New York on March 28, 2025.

Michael M. Santiago | Getty Images News | Getty Images

CoreWeave shares fell about 6% in extended trading on Tuesday even as the provider of artificial intelligence infrastructure beat estimates for second-quarter revenue

Here’s how the company did in comparison with LSEG consensus:

  • Earnings per share: Loss of 21 cents
  • Revenue: $1.21 billion vs. $1.08 billion expected

Revenue more than tripled from $395.4 million a year earlier, CoreWeave said in a statement. The company registered a $290.5 million net loss, compared with a $323 million loss in second quarter of 2024. CoreWeave’s earnings per share figure wasn’t immediately comparable with estimates from LSEG.

CoreWeave’s operating margin shrank to 2% from 20% a year ago due primarily to $145 million in stock-based compensation costs. This is CoreWeave’s second quarter of full financial results as a public company following its IPO in March.

CoreWeave pointed to an expansion in business with OpenAI, a major client and investor. Also during the quarter, CoreWeave acquired Weights and Biases, a startup with software for monitoring AI models, for $1.4 billion.

In May, management touted 420% revenue growth, alongside widening losses and nearly $9 billion in debt. The stock still doubled anyway over the course of the next month.

CoreWeave shares became available on Nasdaq at the end of the first quarter, after the company sold 37.5 shares at $40 each, yielding $1.5 billion in proceeds. As of Tuesday’s close, the stock was trading at $148.75 for a market cap of over $72 billion.

A CoreWeave data center project with up to 250 megawatts of capacity is set to be delivered in 2026, the company said in the statement.

Executives will discuss the results and issue guidance on a conference call starting at 5 p.m. ET.

This is breaking news. Please check back for updates.

WATCH: Citi’s Tyler Radke’s bullish call on CoreWeave, upgraded to buy

Citi's Tyler Radke's bullish call on CoreWeave, upgraded to buy

Continue Reading

Technology

White House says it’s working out legality of Nvidia and AMD China chip deals

Published

on

By

White House says it's working out legality of Nvidia and AMD China chip deals

U.S. President Donald Trump (L) invites Nvidia CEO Jensen Huang to speak in the Cross Hall of the White House during an event on “Investing in America” on April 30, 2025 in Washington, DC.

Andrew Harnik | Getty Images

The Trump administration is still working out the details of its 15% export tax on Nvidia and AMD and could bring deals of this kind to more companies, the White House’s Karoline Leavitt said Tuesday.

“Right now it stands with these two companies. Perhaps it could expand in the future to other companies,” said Leavitt, the White House’s spokesperson.

“The legality of it, the mechanics of it, is still being ironed out by the Department of Commerce, and I would defer you to them for any further details on how it will actually be implemented,” she continued.

President Donald Trump confirmed on Monday that he had negotiated a deal with Nvidia in which the U.S. government approves export licenses for the China-specific H20 AI chip in exchange for a 15% cut of revenue. Advanced Micro Devices also got licenses approved in exchange for a proportion of its China sales, the White House confirmed.

“I said, ‘If I’m going to do that, I want you to pay us as a country something, because I’m giving you a release,'” Trump said Monday.

“We follow rules the U.S. government sets for our participation in worldwide markets,” Nvidia said in a statement this week.

Trump said the export licenses for AMD and Nvidia were a done deal. But lawyers and experts who follow trade have warned that Trump’s deal may be complicated because of existing laws that regulate how the government can charge fees for export licenses.

The Commerce Department didn’t immediately return a request for comment.

The H20 is Nvidia’s Chinese-specific chip that is slowed down on purpose to comply with U.S. export relations. It’s related to the H100 and H200 chips that are used in the U.S., and was introduced after the Biden administration implemented export controls on artificial intelligence chips in 2023.

Earlier this year, Nvidia said that it was on track to sell more than $8 billion worth of H20 chips in a single quarter before the Trump administration in April said that it would require a license to export the chip.

Trump signaled in July that he was likely to approve export licenses for the chip after Nvidia CEO Jensen Huang visited the White House.

The U.S. regulates AI chips like those made by Nvidia for national security reasons, saying that they could be used by the Chinese government to leapfrog U.S. capabilities in AI, or they could be used by the Chinese military or linked groups.

The Chinese government has been encouraging local companies in recent weeks to avoid using Nvidia’s H20 chips for any government or national security-related work, Bloomberg reported on Tuesday.

WATCH: Access to Nvidia’s H20 won’t hand China an AI advantage: Analyst

Access to Nvidia's H20 won't hand China an AI advantage: Analyst

Continue Reading

Trending