Announcing Latitude.sh Launchpad
Today we are excited to announce that Latitude.sh Launchpad is now in public beta. Launchpad is a container-based GPU cloud for model fine-tuning and inference.
Our goal is to do as much of the heavy lifting as possible to give developers the tools they need to build on and serve open-source AI.
We provision the containers directly on bare metal, which means your container's resources are fully dedicated. All containers are powered by 16 CPUs, 185 GB of RAM, 1 TB of NVMe Storage, and a single GPU, which you can choose from the two options available in our beta:
NVIDIA's H100 (80 GB vRAM)
NVIDIA's L40S (48 GB vRAM)
Multi-GPU containers and more GPU options will be available soon. Check out the docs →
NVIDIA's L40S is Now Available
We are excited to introduce a new GPU option to our products, NVIDIA's L40S. This new chip from NVIDIA is powered by the Ada Lovelace Architecture, designed with 4th Generation Tensor Cores, 3rd Generation Ray-Tracing Cores, and CUDA cores, bringing 91.6 Teraflops of FP32 tensor compute.
With 48 GB of vRAM, this new GPU is a great option for running inference and has enough memory to train models with up to 24 Billion parameters in FP16.
The L40S is now available for Latitude.sh Accelerate and Latitude.sh Launchpad.
Accelerate - From $1.04/hr
8 x NVIDIA L40S
Dual AMD 9354, 64 Cores @ 3.25 GHz
1536 GB RAM
4 x 3.8TB NVME
2 x 10 Gbps NIC
Launchpad - From $1.32/hr
1x NVIDIA L40S
16 CPUs
185 GB of RAM
1 TB of NVMe Storage
Introducing lsh, the Latitude.sh CLI
The Latitude.sh command-line interface, lsh, is now available!
The CLI brings Latitude.sh straight to your terminal: from auditing to provisioning, lsh makes it easy to retrieve data about your infrastructure and perform a variety of key actions.
With a simple command, you can quickly check out available plans, get information about your servers, assign a server to a virtual network, and scale your infrastructure up or down by adding or deleting servers, projects, API keys, SSH keys and VLANs. Getting started with the CLI is as easy as installing it and adding your API token. For more information about it, including the full list of commands and examples, click on the link below.
Our CLI is fully open source and we'll improve it based on your contributions. Go to the project repo and give us your feedback.
Improvements & fixes
API version 2023-06-01: Version 2023-06-01 of the API is now stable. Version 2022-07-18 has been marked as deprecated and support will be removed in June 2024. This API introduces breaking changes. Please review them on the API changelog.
Dashboard updates: The sidebar has been updated to provide a simplified navigation experience. Pages Metal and Accelerate have been merged into a single page called Servers. This page now has a native filter called GPU Servers.
User invite email: Invited users will now receive an email when they are added to a team.
Private Management IPs: The private IPv4 IPs that come with your instance don't have a purpose unless you are using Remote Access. They have been removed from API responses and the dashboard to make the information available more relevant to your day-to-day operations.
GitHub authentication: We have added the ability to sign in using your GitHub account.
Save on your bill with On Demand Monthly
We announced hourly billing earlier this year to remove the friction of getting started with Latitude.sh. It is great when you want to run PoCs or to quickly scale up and down. For predictable workloads, it is often better to have consistent, linear pricing for your resources.
Today, we are adding the ability for you to choose between hourly and monthly billing when deploying a new server. With monthly billing you save 5% compared to the hourly rate.
The same project can have both hourly and monthly servers. You choose how to be billed at the server level, giving you a lot of flexibility when building your infrastructure.