Cerebras Beats NVIDIA Blackwell in Llama 4 Maverick Inference
SUNNYVALE, Calif., May 28, 2025--(BUSINESS WIRE)--Last week, Nvidia announced that 8 Blackwell GPUs in a DGX B200 could demonstrate 1,000 tokens per second (TPS) per user on Meta's Llama 4 Maverick. Today, the same independent benchmark firm Artificial Analysis measured Cerebras at more than 2,500 TPS/user, more than doubling the performance of Nvidia's flagship solution.
"Cerebras has beaten the Llama 4 Maverick inference speed record set by NVIDIA last week," said Micah Hill-Smith, Co-Founder and CEO of Artificial Analysis. "Artificial Analysis has benchmarked Cerebras' Llama 4 Maverick endpoint at 2,522 tokens per second, compared to NVIDIA Blackwell's 1,038 tokens per second for the same model. We've tested dozens of vendors, and Cerebras is the only inference solution that outperforms Blackwell for Meta's flagship model."
With today's results, Cerebras has set a world record for LLM inference speed on the 400B parameter Llama 4 Maverick model, the largest and most powerful in the Llama 4 family. Artificial Analysis tested multiple other vendors, and the results were as follows: SambaNova 794 t/s, Amazon 290 t/s, Groq 549 t/s, Google 125 t/s, and Microsoft Azure 54 t/s.
Andrew Feldman, CEO of Cerebras Systems, said, "The most important AI applications being deployed in enterprise today—agents, code generation, and complex reasoning—are bottlenecked by inference latency. These use cases often involve multi-step chains of thought or large-scale retrieval and planning, with generation speeds as low as 100 tokens per second on GPUs, causing wait times of minutes and making production deployment impractical. Cerebras has led the charge in redefining inference performance across models like Llama, DeepSeek, and Qwen, regularly delivering over 2,500 TPS/user."
With its world record performance, Cerebras is the optimal solution for Llama 4 in any deployment scenario. Not only is Cerebras Inference the first and only API to break the 2,500 TPS/user milestone on this model, but unlike the Nvidia Blackwell used in the Artificial Analysis benchmark, the Cerebras hardware and API are available now. Nvidia used custom software optimizations that are not available to most users. Interestingly, none of the Nvidia's inference providers offer a service at Nvidia's published performance. This suggests that in order to achieve 1000 TPS/user, Nvidia was forced to reduce throughput by going to batch size 1 or 2, leaving the GPUs at less than 1% utilization. Cerebras, on the other hand, achieved this record-breaking performance without any special kernel optimizations, and it will be available to everyone through Meta's API service coming soon.
For cutting-edge AI applications such as reasoning, voice, and agentic workflows, speed is paramount. These AI applications gain intelligence by processing more tokens during the inference process. This can also make them slow and force customers to wait. And when customers are forced to wait, they leave and go to competitors who provide answers faster—a finding Google showed with search more than a decade ago.
With record-breaking performance, Cerebras hardware and resulting API service is the best choice for developers and enterprise AI users around the world.
For more information, please visit https://www.cerebras.ai/.
View source version on businesswire.com: https://www.businesswire.com/news/home/20250528123694/en/
Contacts
pr@zmcommunications.com
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Boston Globe
20 minutes ago
- Boston Globe
Amazon's emissions climbed 6% in 2024 on data center buildout
Get Starting Point A guide through the most important stories of the morning, delivered Monday through Friday. Enter Email Sign Up TECH Advertisement Nvidia CEO downplays role in lifting US ban on chip sales to China CEO of Nvidia Jensen Huang spoke to journalists during a press conference at the Mandarin Oriental Qianmen after attending the third China International Supply Chain Expo, in Beijing, on Wednesday. Andy Wong/Associated Press The head of Nvidia downplayed his role in getting the US government to lift a ban on selling an advanced computer chip in China and said it will take time to ramp up production once orders for the AI-processor come in. CEO Jensen Huang, speaking Wednesday in the Chinese capital Beijing, was upbeat about the prospects for the H20 chip, which was designed to meet US restrictions on technology exports to China but nonetheless blocked in April. He met President Trump before his trip and his company announced this week it had received assurances that sales to China would be approved. 'I don't think I changed his mind,' Huang told a cluster of journalists, many of whom asked for his autograph or to take selfies with him. He described his role as informing governments in the United States and elsewhere of the nature and unintended consequences of their policies. The decision to lift the ban on the H20 chip was entirely in the hands of the American and Chinese governments and whatever trade talks they had, he said. 'We can only influence them, inform them, do our best to provide them with facts,' Huang said. 'And then beyond that is out of our control.' — ASSOCIATED PRESS Advertisement ECONOMY Fed Beige Book shows slight improvement People walk in a shopping district along 5th Avenue in Manhattan on July 14. Spencer Platt/Getty US economic activity 'increased slightly' between late May and early July, the Federal Reserve said in its Beige Book survey of regional business contacts. 'That represented an improvement over the previous report, in which half of districts reported at least slight declines in activity,' according to the report published Wednesday. 'Uncertainty remained elevated, contributing to ongoing caution by businesses.' All 12 regions of the country reported price increases, with businesses experiencing 'modest to pronounced input cost pressures related to tariffs,' the Beige Book said. 'Many firms passed on at least a portion of cost increases to consumers through price hikes or surcharges, although some held off raising prices because of customers' growing price sensitivity, resulting in compressed profit margins,' according to the report. The Boston Fed compiled the latest edition of the Beige Book using information gathered on or before July 7. The report includes commentary and anecdotes from business leaders and other contacts in each of the Fed's 12 regional districts. Fed officials next meet July 29-30. — BLOOMBERG NEWS Advertisement PHARMACEUTICALS J&J shrugs off Trump tariff threat and boosts outlook The Johnson & Johnson campus in New Brunswick, N.J. Mark Kauzlarich/Bloomberg Johnson & Johnson beat Wall Street's quarterly sales expectations and raised its full-year outlook, a show of confidence as the pharmaceutical industry faces the dual threats of tariffs and a crackdown on drug pricing. J&J's strong second quarter comes as President Trump floats the idea of levies on the sector. On Tuesday night, he said tariffs on drugs could 'probably' come at the end of the month, starting low and working their way up. A week earlier, Trump told reporters he would impose tariffs as high as 200 percent on drug companies if they don't shift more of their manufacturing to the US over the next year to 18 months. A slow ramp-up to tariffs would actually be good news, J&J chief financial officer Joe Wolk said in an interview. The delay shows 'there's an understanding you can't put up a biopharmaceutical manufacturing facility overnight,' Wolk said. 'As long as those conversations continue to occur, I think we're in a pretty good position.' Shares rose 6.2 percent in Wednesday trading. J&J is often an industry bellwether as the first heath care company to report earnings each quarter. Its performance is being closely watched as drugmakers operate under the specter of potential tariffs and a new policy that seeks to make US drug prices among the lowest in the world. — BLOOMBERG NEWS HEALTH Kennedy fires 2 top aides in department shake-up US Health and Human Services (HHS) Secretary Robert F. Kennedy Jr. a roundtable discussion on soil health at the US Capitol on July 15. Michael M. Santiago/Getty Health Secretary Robert F. Kennedy Jr. fired two of his top aides this week, shaking up his leadership team at the Department of Health and Human Services amid an aggressive effort to reshape public health policy. The firings were of Heather Flick Melanson, his chief of staff, and Hannah Anderson, his deputy chief of staff for policy, according to two people familiar with the matter. They spoke on the condition of anonymity because they were not authorized to speak publicly about the personnel changes. The reason for the ousters, which were first reported by CNN, was not immediately clear. Both Flick Melanson and Anderson were seen as steady and effective government veterans. During a recent Senate hearing, Kennedy indicated that they were the people in his office to call to get things done. When one lawmaker raised a concern about regulations governing the amount of sugar in orange juice, Kennedy advised: 'Why don't you call Heather Flick or Hannah Anderson this week? And we will act on that as quickly as we can.' But Kennedy has also demonstrated that he is willing to muscle his agenda forward. He has pushed ahead with a plan to fire about 20,000 staff members from the health department and pare down 28 divisions to 15 — all amid court challenges. — NEW YORK TIMES Advertisement
Yahoo
34 minutes ago
- Yahoo
Apollo economist warns: AI bubble now bigger than 1990s tech mania
-- Apollo Global Chief Economist Torsten Sløk is sounding the alarm on what he sees as an even more inflated market than the one that led to the dot-com crash. In a striking warning today, Sløk said, 'The difference between the IT bubble in the 1990s and the AI bubble today is that the top 10 companies in the S&P 500 today are more overvalued than they were in the 1990s.' His comments come as stocks continue to reach new highs, with investors increasingly betting big on artificial intelligence as the next transformative force in global markets. The enthusiasm has driven tech giants like NVIDIA (NASDAQ:NVDA), Microsoft Corporation (NASDAQ:MSFT), and Meta (NASDAQ:META), among others, to record-breaking valuations, with NVIDIA recently becoming the first company to trade over the $4 trillion valuation level. Slok highlights that the top 10 stocks have disproportionately high valuations relative to the overall market. Sløk argues that this extreme concentration and the sky-high expectations built into prices mirror and surpass the late-1990s mania, when investors poured money into internet stocks with little regard for profitability. Unlike the IT bubble, today's top companies are highly profitable, but Sløk cautions that even strong fundamentals can't justify unlimited multiples. As AI fever continues to spread from Silicon Valley to Wall Street, Sløk's warning serves as a sobering counterpoint to the hype. Whether his fears prove prophetic or overly cautious, his message is clear: this is not the 1990s but potentially something bigger, and riskier. Speaking of AI, read this: Surge of 50% since our AI selection, this chip giant still has great potential Related articles Apollo economist warns: AI bubble now bigger than 1990s tech mania Surge of 50% since our AI selection, this chip giant still has great potential Apollo economist warns: AI bubble now bigger than 1990s tech mania
Yahoo
34 minutes ago
- Yahoo
625 student beds underway in high-demand University of Michigan market
This story was originally published on Multifamily Dive. To receive daily news and insights, subscribe to our free daily Multifamily Dive newsletter. Property: Chapter Ann Arbor Developers: CRG, Shapack Partners Architect: Lamar Johnson Collaborative Location: Ann Arbor, Michigan Units: 183 (625 student beds) Cost: Withheld Chicago-based real estate developers CRG and Shapack Partners have obtained full construction financing and closed on the future site of a 19-story student housing property in Ann Arbor, Michigan's South University District, near the central campus of the University of Michigan, according to a press release shared with Multifamily Dive. Site work for Chapter Ann Arbor is set to begin this month, with a formal groundbreaking ceremony planned for August. Delivery is expected in time for the start of the 2027-28 academic year. Chapter Ann Arbor is intended to meet the demand for student housing in Ann Arbor, as enrollment at the University of Michigan is outpacing the housing supply in the area, according to the release. The property will offer 625 student beds across 183 units, ranging from studios to five-bedroom apartments. It will be located one block from the campus, within a short walking distance of the Diag central green, according to J.J. Smith, executive vice president and partner at CRG. A variety of shops, restaurants and entertainment venues are located nearby on South University Avenue. 'Chapter Ann Arbor is a strategic addition to our national portfolio, addressing the growing demand for high-quality housing near top-tier universities,' Smith said in the release. The modern-style property was designed by Chicago-based architect Lamar Johnson Collaborative, a subsidiary of Clayco, CRG's parent company. Each unit will be fully furnished, and will include wood-style flooring, quartz countertops, stainless steel appliances, smart TVs and in-unit laundry. Full-building amenities will include a fitness center with a wellness studio and sauna, study lounges, a library, private collaboration rooms and a rooftop patio with fire pits, grilling stations and a hot tub. The University of Michigan's supply of on-campus student beds has remained the same even as enrollment has increased, according to The Michigan Daily. The school does not guarantee housing for undergraduates after their first year, and only offers 1,100 beds for returning students — for which 2,800 students applied in fall 2024. The university recently partnered with Bee Cave, Texas-based student housing developer American Campus Communities to add 2,300 dormitory-style beds to its campus, expected to be completed in 2026. Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
