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.
Cisco CEO Chuck Robbins speaks at the Business Roundtable CEO Workforce Forum in Washington on June 17, 2025.
Al Drago | Bloomberg | Getty Images
CIsco reported results on Wednesday that narrowly exceeded analysts’ expectations and issued quarterly guidance that was also better than expected. The stock slipped in extended trading.
Here’s how the company did in comparison with LSEG consensus:
Earnings per share: 99 cents adjusted vs. 98 cents expected
Revenue: $14.67 billion vs. $14.62 billion expected
Revenue increased 7.6% year over year in the quarter, which ended on July 26, according to a statement. Net income rose to $2.82 billion, or 71 cents per share, from $2.16 billion, or 54 cents per share, in the same quarter a year ago.
Management called for 97 cents to 99 cents in fiscal firsœt-quarter adjusted earnings per share on $14.65 billion to $14.85 billion in revenue. Analysts surveyed by LSEG were expecting 97 cents per share on $14.62 billion in revenue.
For the full 2026 fiscal year, Cisco forecast $4 to $4.06 in adjusted earnings per share and $59 billion to $60 billion in revenue. The LSEG consensus was for earnings of $4.03 a share and $59.53 billion in revenue.
“While we have some clarity on tariffs, we are still operating in a complex environment,” Mark Patterson, Cisco’s finance chief, said on a conference call with analysts.
In the fiscal fourth quarter, Cisco generated $7.63 billion in networking revenue, up 12%. Analysts polled by StreetAccount were looking for $7.34 billion.
Cisco’s security revenue for the quarter totaled $1.95 billion, up 9% and trailing the StreetAccount estimate of $2.11 billion.
During the quarter, Cisco said it would collaborate with a partnership to invest in artificial intelligence infrastructure, alongside BlackRock, Microsoft and other companies. It joined a Stargate data center initiative for the Middle East that involves OpenAI and SoftBank. And the company introduced switches and routers that can take on AI workloads.
AI infrastructure orders from web companies in the quarter reached $800 million, Cisco CEO Chuck Robbins said on the call. The total for the 2025 fiscal year was over $2 billion, more than double the company’s goal, he said.
Cisco’s AI infrastructure sales pipeline from enterprises is in the hundreds of billions of dollars, Robbins said.
At market close on Wednesday, Cisco shares are up 19% in 2025, while the S&P 500 has gained about 10%.
This is breaking news. Please check back for updates.
In its second quarterly financial results as a public company, CoreWeave reported an adjusted loss of 27 cents per share, compared to a 21-cent loss per share expected by analysts polled by LSEG.
CoreWeave’s results came as the lock-up period following its initial public offering is set to expire Thursday evening and potentially add volatility to shares. The term refers to a set period of time following a market debut when insiders are restricted from selling shares.
“We remain constructive long term and are encouraged by today’s data points, but see near-term upside capped by the potential CORZ related dilution and uncertainty, and the pending lock-up expiration on Thursday,” wrote analysts at Stifel, referencing the recent acquisition of Core Scientific.
Shares of Core Scientific fell 7% Wednesday.
In the current quarter, the company projects $1.26 billion to $1.30 billion in revenue. Analysts polled by LSEG forecasted $1.25 billion. CoreWeave also lifted 2025 revenue guidance to between $5.15 billion and $5.35, up from a $4.9 billion to $5.1 billion forecast provided in May and above a $5.05 billion estimate.
Read more CNBC tech news
Some analysts were hoping for stronger guidance given the stock’s massive surge since going public in March. Others highlighted light capital expenditures guidance and a delay in some spending until the fourth quarter as a potential point of weakness.
“This delay in capex highlights the uncertainty around deployment time; as go-live timing is pushed, in-period revenue recognition will be smaller,” wrote analysts at Morgan Stanley.
The AI infrastructure provider said revenue more than tripled from a year ago to $1.21 billion as it continues to benefit from surging AI demand. That also surpassed a $1.08 billion forecast from Wall Street. Finance chief Nitin Agrawal also said during a call with analysts that demand outweighs supply.
The New Jersey-based company, whose customers include OpenAI, Microsoft and Nvidia, also said it has recently signed expansion deals with hyperscale customers.
CoreWeave acquired AI model monitoring startup Weights and Biases for $1.4 billion during the period and said it finished the quarter with a $30.1 billion revenue backlog.
Apple CEO Tim Cook (R) shakes hands with U.S. President Donald Trump during an event in the Oval Office of the White House on August 6, 2025 in Washington, DC.
Win Mcnamee | Getty Images
Top tech executives are at the forefront of a recent swathe of unprecedented deals with U.S. President Donald Trump.
In just the last few days, the White House confirmed that two U.S. chipmakers, Nvidia and Advanced Micro Devices, would be allowed to sell advanced chips to China in exchange for the U.S. government receiving a 15% cut of their revenues in the Asian country.
Apple CEO Tim Cook, meanwhile, recently announced plans to increase the firm’s U.S. investment commitment to $600 billion over the next four years. The move was widely seen as a bid to get the tech giant out of Trump’s crosshairs on tariffs — and appears to have worked for now.
Altogether, analysts say the deals show just how important it is for the world’s largest companies to find some tariff relief.
“The flurry of deal-making is an effort to secure lighter treatment from tariffs,” Paolo Pescatore, technology analyst at PP Foresight, told CNBC by email.
“In some shape or form, all of the big tech companies have been negatively impacted by tariffs. They can ill afford to fork out on millions of dollars in additional fees that will further dent profits as underlined by recent quarterly earnings,” Pescatore said.
While the devil will be in the detail of these agreements, Pescatore said that Apple leading the way with its accelerated U.S. investment will likely trigger “a domino effect” within the industry.
Apple, for its part, has long been regarded as one of the Big Tech firms most vulnerable to simmering trade tensions between the U.S. and China.
Earlier this month, Trump announced plans to impose a 100% tariff on imports of semiconductors and chips, albeit with an exemption for firms that are “building in the United States.”
Apple, which relies on hundreds of different chips for its devices and incurred $800 million in tariff costs in the June quarter, is among the firms exempt from the proposed tariffs.
A ‘hands-on’ approach
The Nvidia and AMD deal with the Trump administration has meanwhile sparked intense debate over the potential impact on the chip giants’ businesses and whether the U.S. government may seek out similar agreements with other firms.
White House spokesperson Karoline Leavitt said Tuesday that the legality and mechanics of the 15% export tax on Nvidia and AMD were “still being ironed out.” She also hinted deals of this kind could expand to other companies in future.
Ray Wang, founder and chairman of Constellation Research, described the Nvidia and AMD deal to pay 15% of China chip sales revenues to the U.S. government as “bizarre.”
Speaking on CNBC’s “Squawk Box” on Monday, Wang said what is “really weird” is there is still some uncertainty over whether these chips represent a national security issue.
“If the answer is no, fine OK. The government is taking a cut out of it,” Wang said. “Both Nvidia’s Jensen Huang and Lisa Su at AMD both decided that OK, we’ve got a way to get our chips into China and maybe there is something good coming out of it.”
Investor concerns
While investors initially welcomed the deal as broadly positive for both Nvidia and AMD, which once more secure access to the Chinese market, Wang said some in the industry will nevertheless be concerned.
“As an investor, you’re worried because then, is this an arbitrary decision by the government? Does every president get to play kingmaker in terms of these deals?” Wang said.
“So, I think that’s really what the concern is, and we still have additional tariffs and trade deals to come from the China negotiations,” he added.
Looking ahead, Dan Niles, founder and portfolio manager at Niles Investment Management, said the question for investors is whether the Trump administration’s “hands-on” approach is positive or negative for U.S. companies.
“I think for each company, it is very different. So, it certainly it is something I take into account. The bigger thing for me is do you have some stability of policy? Do you have a policy one week and then it flips the next?” Niles told CNBC’s “Closing Bell: Overtime” on Monday. “Right now, that is what concerns me a little bit more.”
— CNBC’s Arjun Kharpal and Kif Leswing contributed to this report.