Register and share your invite link to earn from video plays and referrals.

Search results for MLPerf
MLPerf community
One keyword maps to one global community path.
Create community
People
Not Found
Tweets including MLPerf
New MLPerf® Storage v3.0 results are in! Nebius Object Storage fed 768 simulated B200 accelerators on RetinaNet, the highest accelerator count of any submission in this round. Get the full breakdown: #MLPerf# #AI# #ObjectStorage#
Show more
Five first-place results in MLPerf® Inference v6.1. 603,023 tokens/sec on DeepSeek R1 at 72 GPUs and preview-category results on the Nebius @nvidia Vera Rubin NVL72. We were one of only two submitters with results on that hardware. Full results: #MLPerf# #MLCommons# #Inference#
Show more
Nvidia’s next AI rack just posted its first official MLPerf scores. Vera Rubin NVL72 (preview) vs today’s GB300 rack: • up to 3.7× on Qwen3-VL • up to 2.5× on DeepSeek-R1
Show more
🚨 $NBIS just posted five first-place results in the latest MLPerf AI inference benchmarks, including leading full-rack performance on NVIDIA's GB300 NVL72. Nebius was also one of only two companies to submit results on NVIDIA's next-gen Vera Rubin NVL72 platform, already running DeepSeek R1 at 603K tokens/sec across 72 GPUs in the preview category. Another strong validation of $NBIS' infrastructure, technical execution, and close engineering relationship with NVIDIA.
Show more
tinygrad finally has a SOTA speed result! This is MLPerf llama31_8b in 2h 6m, beating the 2h 7m time from AMD's docker on our computer. We aren't ahead in wall time yet because it's warm in San Diego right now and we are too cheap to buy AC. Can you spot the thermal throttle?
Show more
2.02 minutes. That's the time-to-train for DeepSeek-V3 671B on 8,192 @NVIDIA GB300 NVL72 GPUs in MLPerf® Training v6.0, the fastest result in the round on this benchmark. Our new whitepaper covers the engineering behind the number.
Show more
A TON OF THINGS HAPPENED IN THE STOCK MARKET TODAY. Here's a full recap: 1. The Federal Reserve raised interest rates by 25 bps for the first time since July 2023 in a unanimous 12-0 decision. In its statement, the Fed said Middle East conflicts and broader geopolitical developments contributed to the decision. The updated dot plot now points to one more 25 bps hike in 2026, with 12 of 18 officials expecting exactly one additional hike, 4 expecting two more hikes, and 2 expecting no further hikes. The Fed also projects 2 additional rate hikes in 2027, signaling policymakers expect rates to remain higher for longer. 2. President Trump is expected to meet with Gulf leaders on the sidelines of the U.N. General Assembly in New York next Tuesday to discuss the next phase of the Iran conflict, according to reports. The discussions are expected to focus on U.S. proposals for a postwar strategy, with leaders from the Gulf Cooperation Council (GCC), Saudi Arabia, the UAE, Qatar, Bahrain, Kuwait, and Oman, expected to participate. The meeting comes as the administration works on a broader postwar framework ahead of the Sept. 24 Trump–Xi summit, where regional security and AI are also expected to be discussed. 3. Generac $GNRC signed a long-term agreement to supply backup power generators for Amazon $AMZN data centers, with total payments under the deal potentially reaching $8B. Initial generator deliveries are expected to total roughly $2.4B across 2027 and 2028. As part of the agreement, Amazon received warrants to purchase up to 1.69M Generac shares at $200.93 each, with about 308K shares vesting immediately and the rest vesting in stages as Amazon’s payments move toward the $8B threshold. The warrants are exercisable through September 2033. The deal gives Generac a major long-term hyperscaler customer as AI data center growth continues driving demand for backup power and broader energy infrastructure. 4. Fed Chair Kevin Warsh said today’s rate hike removed a “dose of accommodation,” adding that the move shows the Fed is serious about restoring price stability. Warsh also said he believes the Fed can bring inflation back under control without damaging the labor market: “We don’t need to do harm to the job market to achieve our objective.” On AI, he said the Fed is closely monitoring developments but that decisions around AI’s risks and rewards belong to other policymakers, adding that the Fed’s AI task force is expected to report back by year-end. Meanwhile, White House official Desai called the Fed’s rate hike “rather unfortunate.” 5. Nebius $NBIS announced another ~20% price increase for on-demand H100, H200, B200, and B300 GPUs, effective October 1, just 4 months after raising prices by roughly 30%. The company also posted 5 first-place results in the latest MLPerf AI inference benchmarks, including the top full-rack performance on NVIDIA’s GB300 NVL72. Nebius was one of only 2 companies to submit results on NVIDIA’s next-generation Vera Rubin NVL72 platform, achieving 603K tokens/sec running DeepSeek R1 across 72 GPUs in the preview category. The results further reinforce Nebius’ infrastructure capabilities, engineering execution, and close partnership with NVIDIA. 6. The top 10 most active options today by contracts traded were $NVDA with 3.0M contracts, $TSLA with 2.4M contracts, $AAPL with 1.6M contracts, $SPCX with 1.2M contracts, $INTC with 1.0M contracts, $MU with 857K contracts, $META with 857K contracts, $AMZN with 854K contracts, $AMD with 743K contracts, and $HOOD with 403K contracts. 7. Trump administration officials are reportedly considering a meeting with major AI and tech executives on the sidelines of Xi Jinping’s Washington visit next week, though nothing has been finalized, according to CNN. OpenAI CEO Sam Altman and Nvidia $NVDA CEO Jensen Huang are among those expected to be in Washington for the state dinner and could potentially attend. The proposed discussion would focus on the future of AI and related risks, although Xi is not currently expected to participate. The meeting comes as the U.S. and China also prepare to discuss shared AI risks ahead of the September 24 Trump–Xi summit. 8. Chipotle $CMG is partnering with Palantir $PLTR to strengthen food safety monitoring across its restaurants. Chipotle will use Palantir’s platform to track potential food safety risks, including pest incidents, employee illnesses, and other operational factors that could impact restaurant safety. The partnership comes after a challenging summer for food safety across the U.S. and highlights how Palantir continues expanding beyond government and industrial customers into large-scale commercial use cases, where its software is being used to improve operational decision-making and risk management. 9. U.S. homebuilder confidence fell to 32, its weakest reading since late 2022, as elevated mortgage rates continued weighing on demand. The average 30-year mortgage rate has climbed back above 7%, near its highest level in almost 2 years. Builders are increasingly responding with incentives, with 38% cutting prices and 66% offering buyer concessions. At the same time, housing inventory has reached a 7-year high, while Redfin estimates there were 58% more sellers than buyers in August, the widest imbalance since it began tracking the data in 2013. 10. Nokia $NOK is expanding AI-RAN trials with operators across North America, Europe, Asia-Pacific, and the Middle East as carriers move from early evaluations into lab and live-field testing. The platform combines Nokia’s anyRAN software with NVIDIA’s Aerial RAN Computer and has delivered more than 20% improvement in spectral efficiency, according to Nokia. New and ongoing partners include A1 Group, Chunghwa Telecom, du, e&, Mobily, stc, TPG Telecom, Zain Saudi, and NTT DOCOMO, while earlier evaluations involved T-Mobile, SoftBank, and Indosat. Nokia and NVIDIA are also expanding their AI-RAN collaboration as the telecom industry moves toward AI-native 5G and 6G networks. 11. Apple $AAPL is reportedly weighing a return to the server market with an AI-focused enterprise system built around its own M-series chips and NVIDIA’s NVLink Fusion networking technology, according to The Information. The proposed servers would use either 2 or 4 M8 Ultra chips and target AI developers, enterprises, and governments looking to run AI inference on private infrastructure. The project is currently aimed for 2029, though the plans could still change or be canceled. The move follows surging demand from AI developers for Mac mini and Mac Studio systems, helping drive Mac revenue up nearly 29% last quarter to $10.4B. If launched, it would mark Apple’s return to the server market for the first time since discontinuing Xserve in 2011. 12. CoreWeave $CRWV signed a definitive 15-year anchor lease with Blockfusion for capacity at its Niagara Falls, New York AI data center campus. The former power plant is being converted into a high-density, liquid-cooled AI facility powered predominantly by hydroelectric energy. The agreement further expands CoreWeave’s long-term AI infrastructure footprint while leveraging low-cost, renewable power to support next-generation GPU deployments. WALL STREET IS THE GREATEST SHOW ON EARTH.
Show more