Nebius Group

We release world’s first datasets for training software engineering agents🔥 More specifically, our AI R&D team uploaded two datasets to Hugging Face: nebius/SWE-bench-extra, containing 6,411 Issue-Pull Request pairs, and nebius/SWE-agent-trajectories, featuring 80,036 software engineering agent trajectories, where an agent attempts to solve these issues. To learn more about this release, read our today’s research blog post: - https://lnkd.in/dg_XuRJt Check out the datasets on HF: - https://lnkd.in/eDDB94re - https://lnkd.in/eMbcA5Np #HuggingFace #agents #datasets #release

It has been an amazing year for Nebius Group, and as we head into the holiday season, we’ve been reflecting on what we’ve achieved in 2024. This year we launched as a group, but it has been so much more than that. It’s hard to capture everything in one post, but if we had to choose a few key highlights:    💹 Restarting trading on Nasdaq – our shares restarted trading in October under our new ticker symbol “NBIS”.  💰 Securing USD 700M – in December we announced a strategic equity financing with investors including NVIDIA, Accel and Orbis Investments to accelerate our full-stack AI infrastructure rollout. 🚀 Launching our AI-native cloud – in October Nebius unveiled its new cloud computing platform, built from scratch for the age of AI to manage the full ML lifecycle all in one place. 🤝 Avride x Uber – also in October, our self-driving business partnered with Uber, bringing Avride’s delivery robots and vehicles to Uber and Uber Eats.    🏆 TripleTen scoops bootcamp award – Fortune Magazine in May named TripleTen as the best software engineering boot camp in the US to teach total beginners the language and tools needed to become a full-stack developer.    💫 Toloka moves into GenAI – Our data business evolved to become a trusted partner for leading global foundation model producers and Big Tech companies. Thank you to all of our colleagues, friends and partners who have been on this journey with us so far. See you in 2025 to continue to build on this year’s successes!  

TractoAI was built by Nebius as a serverless GPU cloud platform to enable AI developers to focus on training and fine-tuning AI models. Pleias, a leading French AI research lab, partnered with TractoAI to train their 1B LLM. Intrigued? Read the full story 🔗 https://lnkd.in/eBsRZpUE

Avride’s robots have officially hit the streets of downtown Dallas, and are now delivering Uber Eats orders! The partnership with Uber Eats allows Avride to bring reliable, efficient, and innovative delivery solutions to customers in the heart of the city – the first time delivery robots have operated in downtown Dallas. Read more 🔗 https://lnkd.in/e2FjE_M2

AI's potential is vast, but ensuring its ethical and helpful behavior requires proper alignment. In a thought-provoking article for Forbes, Toloka's CEO Olga Megorskaya, discusses how human feedback plays a vital role in model alignment. “Alignment is a key piece of the responsible AI puzzle. Better alignment with ethical standards will translate to stronger trust in AI systems and increased user adoption," Olga writes. "When we rally behind responsible AI, we have the potential to energize AI safety efforts across the industry. Together, we can pursue deeper exploration into model alignment and fine-tune the ethics of AI” Read the full article to learn more about the critical role of human preference in shaping the future of AI 🔗 https://lnkd.in/e8GD894A

How will AI reshape our world?     Our CEO and founder, Arkady Volozh, recently shared his perspective on the transformative role AI will play in our lives and across industries at the recent FT Live Future of AI Summit in London.     Dive into his insights from this fireside chat 👉https://lnkd.in/eZS8SHY4

Introducing U-MATH: A comprehensive LLM benchmark for university-level math  ✨ We compared small, large, and proprietary models — and the leader is not GPT-4o. ✨ Toloka and Gradarius have joined forces to build the U-MATH benchmark, challenging LLMs with math problems from curriculum at top US universities. Our testing reveals that even top-tier models fall short in practical math skills, but the leader in university math is Gemini 1.5 Pro. The U-MATH benchmark has 1,100 math problems across 6 university-level math subjects, with 20% of the problems requiring visual comprehension — the only benchmark of its kind in the industry.  It’s accompanied by a meta-eval benchmark, μ-MATH. For insights into math performance and details on how U-MATH was collected, read our post: https://bit.ly/4f4Cg7g Want to check your model’s performance? Download from Hugging Face — links in the comments! 👇 Or connect with our team to discuss in-depth evaluation and fine-tuning with our math dataset. #GenAI #LLMs #Benchmarks #LLMeval #mathAI #GeminiPro

We are excited to announce that Nebius will deploy NVIDIA’s next-generation Blackwell GPUs at data centers in the US and Finland. Clusters of NVIDIA GB200 NVL72 Grace Blackwell Superchips and NVIDIA HGX B200 systems will be available to customers of Nebius’s AI-native cloud from early 2025. By offering these solutions in Europe and the US, Nebius eliminates latency and gives customers from scale-ups to enterprises a seamless infrastructure for building, training and inferencing large-scale AI models. Read more 🔗 https://lnkd.in/dw6RuhxF

Today, we’re excited to share our investment in Nebius, a full-stack AI platform built for intensive AI & ML workloads. As AI reshapes industries and drives innovation, reliable, scalable infrastructure has become paramount. Nebius, with its years-long expertise in building one of the top tech infrastructures, delivers a vertically integrated AI-native solution optimized from data center to cloud platform. We’ve been fortunate to know the Nebius team since its early days, and we’re thrilled to partner with founder Arkady Volozh and the entire team in fueling this era of rapid AI innovation. Read more from our partner, Matt Weigand, on today’s news #AccelFamily https://lnkd.in/gHr52M2t

We’re thrilled to announce a USD 700 million strategic investment from investors including Accel, NVIDIA and Orbis Investments. The investment supports our previously announced plans to further build out our full-stack AI infrastructure – including large-scale GPU clusters, cloud platforms and tools and services for developers – for AI pioneers globally. In the words of our founder and CEO, Arkady Volozh: “The foundation of our business is our expertise in building advanced technology infrastructure. We have demonstrated the scale of our ambitions, initiating an AI infrastructure build-out across two continents. This strategic financing gives us additional firepower to do it faster and on a larger scale. I’m grateful to our investors for the trust they have placed in us – our team is ready to deliver.” Matt Weigand, partner at Accel who led the investment round and will be joining Nebius's Board of Directors, said: "When we first met Arkady and the Nebius team, we knew that we wanted to partner together: their team's unique combination of hardware and software expertise allows Nebius to scale infrastructure from the server racks to the apps. Building their own data centers and AI platform on top, Nebius can optimize every part of the stack to deliver higher quality and lower cost infrastructure, and it is well-equipped to become an important utility of the AI age.” Read the full story 🔗 https://lnkd.in/eydgaqN3