SF Compute
A GPU cloud and compute market that rents bare metal, managed Slurm and VM clusters on demand or by reservation, with an order book that lets customers sell back the capacity they do not use.
Find alternatives to SF Compute
What it does
- GPU cloud
- Model training
Sources
- Pricing
-
Official documentation · 16 Nov 2026
Docs quickstart: "You set the number of nodes (each node runs 1 instance), start time, duration, rate, and the SKU the order pins to." The rate is set per order against an order book, so the model is usage-based; sfcompute.com/pricing redirects to a sign-in page, so no price table was read (blocked, not absent).
- Description
-
Official documentation · 16 Nov 2026
Home page text fetched 2026-11-16 with curl; page quote: "Get VMs, Bare Metal, & managed Slurm from the only cloud designed to let you sublease your reserved capacity when plans change."; specs page: "Deploy on-demand or reserve instances with no long-term contracts."; "Bare metal and managed Slurm for the lowest-level access, and VMs for flexibility."; "No ingress/egress fees"; docs: "San Francisco Compute is a market for very large scale GPU clusters."; "Run a PyTorch job across multiple nodes". The company's other record, Autoresearch, is a separate subdomain product for agents.
- What it does
-
Official documentation · 16 Nov 2026
Home page text fetched 2026-11-16 with curl; page quote: "Get VMs, Bare Metal, & managed Slurm from the only cloud designed to let you sublease your reserved capacity when plans change."; specs page: "Deploy on-demand or reserve instances with no long-term contracts."; "Bare metal and managed Slurm for the lowest-level access, and VMs for flexibility."; "No ingress/egress fees"; docs: "San Francisco Compute is a market for very large scale GPU clusters."; "Run a PyTorch job across multiple nodes". The company's other record, Autoresearch, is a separate subdomain product for agents.
- Sold within
-
Official documentation · 16 Nov 2026
Home page and docs: the Sign up and Go to dashboard routes and a self-serve CLI quickstart ("sf login") buy compute directly: "You set the number of nodes (each node runs 1 instance), start time, duration, rate, and the SKU the order pins to."
- Name
-
Official documentation · 16 Nov 2026
Home page text fetched 2026-11-16 with curl; page quote: "Get VMs, Bare Metal, & managed Slurm from the only cloud designed to let you sublease your reserved capacity when plans change."; specs page: "Deploy on-demand or reserve instances with no long-term contracts."; "Bare metal and managed Slurm for the lowest-level access, and VMs for flexibility."; "No ingress/egress fees"; docs: "San Francisco Compute is a market for very large scale GPU clusters."; "Run a PyTorch job across multiple nodes". The company's other record, Autoresearch, is a separate subdomain product for agents.
- URL
-
Official documentation · 16 Nov 2026
Home page text fetched 2026-11-16 with curl; page quote: "Get VMs, Bare Metal, & managed Slurm from the only cloud designed to let you sublease your reserved capacity when plans change."; specs page: "Deploy on-demand or reserve instances with no long-term contracts."; "Bare metal and managed Slurm for the lowest-level access, and VMs for flexibility."; "No ingress/egress fees"; docs: "San Francisco Compute is a market for very large scale GPU clusters."; "Run a PyTorch job across multiple nodes". The company's other record, Autoresearch, is a separate subdomain product for agents.
- Deployment
-
Official documentation · 16 Nov 2026
Home page text fetched 2026-11-16 with curl; page quote: "Get VMs, Bare Metal, & managed Slurm from the only cloud designed to let you sublease your reserved capacity when plans change."; specs page: "Deploy on-demand or reserve instances with no long-term contracts."; "Bare metal and managed Slurm for the lowest-level access, and VMs for flexibility."; "No ingress/egress fees"; docs: "San Francisco Compute is a market for very large scale GPU clusters."; "Run a PyTorch job across multiple nodes". The company's other record, Autoresearch, is a separate subdomain product for agents.
- Interfaces
-
Official documentation · 16 Nov 2026
Home page text fetched 2026-11-16 with curl; page quote: "Get VMs, Bare Metal, & managed Slurm from the only cloud designed to let you sublease your reserved capacity when plans change."; specs page: "Deploy on-demand or reserve instances with no long-term contracts."; "Bare metal and managed Slurm for the lowest-level access, and VMs for flexibility."; "No ingress/egress fees"; docs: "San Francisco Compute is a market for very large scale GPU clusters."; "Run a PyTorch job across multiple nodes". The company's other record, Autoresearch, is a separate subdomain product for agents.
Also from SF Compute
-
Autoresearch Platform
Autoresearch lets an LLM agent rent GPU nodes by the minute and run batch jobs, sweeps and RL sandboxes on SF Compute's clusters through an MCP server or CLI.
Open SF Compute in the directory
Something wrong here? Send a correction — quote this product id: sf-compute-gpu-cloud.