Nebius AI monthly digest, June 2024

Our June’s highlights included offering up to 8 GPUs via self-service, decoding GPU pricing and adding L40S to our portfolio. We also detailed our Terraform provider, discussed LLM dataprep, announced partnership and introduced feature requests.

Access up to 8 GPUs via self-service right away

Our new customers can now access to up to 8 NVIDIA® GPUs (H100 and others), along with CPU resources, storage and additional services through our user-friendly self-service console. If your fellow colleagues might be interested, share the good news! And here’s the console video walkthrough featuring Cloud Solutions Architect Khamzet Shogenov:

Decoding GPU pricing — tips from a CFO

The GPU cloud market is flourishing, bringing a variety of pricing models. Our CFO Danila Pavlov provides some tips on interpreting price lists, identifying hidden costs and asking your provider the right questions:

NVIDIA L40S: now also available in Nebius AI

Based in our data center in Finland, L40S GPUs support BF16, FP8, INT8 and INT4 precision formats and provide 48 GB of RAM each, making them ideal for inferring models with 8 billion parameters or less.

What’s new on our docs and blog

Explaining Nebius AI provider for Terraform
In addition to the web console and CLI, we now offer a declarative way to manage your resources: define their target states in configuration files and apply them using our Terraform provider. Learn more in the new document dedicated to Terraform and Nebius AI. The solution library maintained by our solution architects on GitHub uses Terraform extensively — check it out for examples.

Data preparation for LLMs: techniques, tools and our established pipeline
Our Senior Software Developer Yury Anapolskiy explored methods and technologies for maximizing efficiency in data collection and preparation for training large models. The article outlines the pipeline in detail and discuss our own chosen workload for dataprep.

Nebius AI and announced technological partnership
We are thrilled to collaborate with a company that develops such a widely used and significant open-source tool as DVC.

Introducing feature requests: share you ideas on advancing our platform
There’s a new, convenient way to share ideas for advancing our AI-centric cloud, discuss them with us and the community and get upvotes from others interested. Introducing the place for requesting Nebius AI features.

Nebius AI team
Sign in to save this post