Nebius AI Studio is now Nebius Token Factory: same platform, new name, more power for running AI at scale.
Nebius monthly digest: March 2025 and NVIDIA GTC
Nebius monthly digest: March 2025 and NVIDIA GTC
NVIDIA GTC was the highlight of last month for us. Our extensive participation included being a platinum sponsor, a strong presence with the booth, two tech talks, interviews with our leaders, partners and customers, and much more.
April 7, 2025
4 mins to read
Announced at NVIDIA GTC: Nebius to offer NVIDIA Blackwell Ultra-powered instances
We will be an early adopter AI cloud provider to offer the new NVIDIA Blackwell Ultra AI factory platform — unlocking the world’s most advanced compute on demand for AI builders and enterprises everywhere to build the next generation of agentic, reasoning and physical AI.
In support of the launch of NVIDIA Blackwell Ultra at GTC 2025, we will give our customers access to NVIDIA GB300 NVL72-powered instances — accelerated by 72 NVIDIA Blackwell Ultra GPUs — by the end of 2025.
Read the press release here and reserve your NVIDIA Blackwell cluster.
Nebius to be an ecosystem partner for NVIDIA Dynamo
NVIDIA Dynamo is an open-source inference serving framework for deploying GenAI in large-scale distributed environments. It provides the most efficient solution for scaling test-time compute. Dynamo on NVIDIA Blackwell boosts throughput on DeepSeek R1 by 30x.
Interviews and tech talks at NVIDIA GTC
- Check out our NVIDIA GTC 2025 coverage featuring Nebius’ leaders, partners such as SemiAnalysis, AI influencers, as well as customers like Higgsfield AI, Captions, KissanAI and CentML.
- We gave two tech talks at the conference: one on how we rebuilt our cloud from scratch in just one year, and the other on enabling the agent-first future from an AI R&D perspective. Click the links to watch!
Recognition by third parties and at own events
-
Nebius received a gold medal in the GPU Cloud ClusterMAX™ Rating System by SemiAnalysis. The company has evaluated us with honesty; there’s still work ahead, but we’re now officially on par with the hyperscalers — even surpassing some of our competitors.
-
We wrapped up our first series of technical meetups with stops in Paris, London and San Francisco — each time providing a deep dive into our cloud, as well as our contributions to AI R&D and open source. You can watch San Francisco sessions on YouTube.
Being busy at conferences doesn’t stop us from bringing new releases
-
JupyterLab is now available in our Standalone Applications service. This fully managed offering eliminates infrastructure complexities, allowing you to focus entirely on your AI/ML projects.
-
Managed MLflow has become generally available — which means it gains more stability, a newer version of the platform, SLAs and a service fee.
-
Nebius and Outerbounds, an AI/ML platform powered by the popular open-source framework Metaflow, are now integrated as part of a broader technology partnership.
-
Our cloud has been added to SkyPilot as well. SkyPilot is an open-source framework that simplifies running AI and batch workloads across cloud platforms.
Let’s not forget about new blog posts and tech docs updates
-
Speeding up compute is critical to push the boundaries of science. Check out how NVIDIA Hopper GPUs performed against CPU workflows to accelerate homology research, deep learning-enabled protein embeddings and ML clustering approaches.
-
In his blog post, our senior Senior ML Solutions Architect Akim Tsvigun demonstrates how to fine-tune an LLM using Nebius AI Studio with a function-calling task as a running example.
-
We’ve finished covering web console’s monitoring dashboards in the documentation. Learn about dashboards and metrics for Object Storage buckets and Managed MLflow clusters.
-
Nebius’s open-source solution for combining Slurm and Kubernetes, Soperator, got its own documentation. Now you can learn more about Soperator, read how to connect to nodes, manage Slurm users, run jobs and monitor them.
-
Read how to use applications in Nebius AI Cloud — a standalone ones and applications for Managed Kubernetes.
-
Learn what notifications Nebius can send you and how to manage them.
-
We expanded networking in Managed Kubernetes with add-ons installed by default, how you can configure them and how to set up the NodeLocal DNSCache feature.
-
Nebius AI Studio docs now explain how to host a model and how to integrate AI Studio with Portkey.
Explore Nebius AI Cloud
Explore Nebius Token Factory
See also
In February, we at Nebius were busy preparing several new data center regions and working on our large-scale participation in NVIDIA GTC. We also rolled out some improvements to the platform and contributions to open source — as well as new initiatives to support AI research: a brand new biotech award and a research credits program.
We have been hard at work this past month, expanding the scope of our materials, which support customer progress on the platform. Our blog has been hopping too, showcasing the latest releases from Nebius AI Studio and the resilient infrastructure potential Nebius AI Cloud, accelerated by NVIDIA, brings for different AI use cases. We wish you a productive Q1, as we empower your AI journey on our platform.
While the holiday season is perhaps not all about numbers — and more about not thinking about them for a while — they still pop up here and there. Whether it’s the GPU capacity you’re planning for next year or how many billions of parameters you’ll use in your next model, numbers are driving us into 2025. Here are some figures we are reflecting on these days — while keeping a sense of fun about them.
Sign in to save this post