01 Compute
Virtual machines with one or eight NVIDIA GPUs are in Private Preview, opening with the first cohort.
Private Preview
GPU Instances page, /compute#instances
01 Compute
General-purpose virtual machines from 2 to 256 vCPU are in Private Preview, opening with the first cohort.
Private Preview
CPU Instances page, /compute#cpu-instances
01 Compute
Multi-node 8-GPU clusters on one InfiniBand fabric are in Private Preview as plain machines, opening with the first cohort.
Private Preview
GPU Clusters page, /clusters
01 Compute
GB300 NVL72 and GB200 NVL72 racks are sold by request, on reserved terms.
By request
Rack-scale page, /compute#rack-scale
01 Compute
GPU capacity at a market price, with an optional max price per GPU-hour, is in Private Preview, opening with the first cohort.
Private Preview
Spot page, /spot
02 Develop
GPU development environments that scale the same Ray code from one GPU to a cluster are in early access.
Early access
Workspaces page, /workspaces
02 Develop
Isolated, short-lived CPU and GPU environments that AI agents create and end through the API are in early access.
Early access
Sandboxes page, /sandboxes
02 Develop
Python workflows that run step by step on Fantasti GPUs are planned.
Planned
Flows page, /flows
03 Data and network
Volumes, shared filesystems and S3-compatible object storage are in Private Preview, opening with the first cohort.
Private Preview
Storage page, /storage
03 Data and network
Private networks, public addresses, security groups, load balancers and private DNS are in Private Preview, opening with the first cohort.
Private Preview
Networking page, /networking
04 Run
Offline inference over a dataset, with checkpoints, retries and a max price, is in Private Preview, opening with the first cohort.
Private Preview
Batch Inference page, /batch-inference
04 Run
Containers that run on a GPU as a job or as an endpoint are in early access.
Early access
Serverless page, /serverless
04 Run
Kubernetes tools, Ray and SkyPilot on your Fantasti cluster are in early access.
Early access
Integrations page, /integrations
Access
One client for the whole platform, as Python, REST and a CLI, is in Private Preview, opening with the first cohort.
Private Preview
API page, /api
Access
A remote MCP server that gives AI agents the Fantasti API as tools is in early access.
Early access
MCP server page, /mcp
Access
One console and one statement across every product are in Private Preview, opening with the first cohort.
Private Preview
Console and one bill page, /pricing#one-bill
05 Operate
The control plane that checks policy, places each workload on one fabric and meters it onto one statement is in Private Preview, opening with the first cohort.
Private Preview
Orchestrator page, /orchestrator
05 Operate
Metrics, logs and traces for everything on Fantasti, with prebuilt dashboards, are in early access.
Early access
Observability page, /observability
05 Operate
Managed MLflow tracking and model registry are in early access.
Early access
MLflow page, /integrations#mlflow
05 Operate
One-click applications installed on your cluster are planned.
Planned
Applications page, /integrations#applications