SF Compute · Rate Limits

Sfcompute Rate Limits

SF Compute does not publish specific numeric request-rate limits for the api.sfcompute.com REST API. Practical constraints are economic rather than request-rate based: order placement is bounded by available account balance, by open market liquidity for the requested instance type, and by the quantity of capacity a buyer can clear. The market-quote endpoint allows long-running requests (the CLI uses a 600s client timeout). Specific per-account RPM/RPS values are not reconciled in this artifact.

Sfcompute Rate Limits is the machine-readable rate-limit profile for SF Compute on the APIs.io network, conforming to the API Commons Rate Limits specification.

It captures 4 rate-limit definitions, measuring requests, seconds, orders, and nodes.

The profile also includes 2 backoff/retry policies defined and response codes documented for throttled.

Tagged areas include GPU, Compute, Marketplace, H100, and Spot Pricing.

4 Limits Throttle: 429
GPUComputeMarketplaceH100Spot PricingRate LimitingQuotasThrottling

Limits

API Requests account
requests
see provider documentation
No published numeric REST request-rate limit; assume standard fair-use throttling.
Quote Request Duration request
seconds
long-running (CLI uses a 600s client timeout)
Market quote requests may block while the orderbook is searched.
Order Placement account
orders
bounded by available balance and market liquidity
Buy orders only fill against open sell-side liquidity for the instance type.
Capacity Quantity account
nodes
bounded by market availability
Available H100/H200 quantity floats with seller supply.

Policies

Economic Limiting
Spend is gated by account balance and orderbook liquidity rather than fixed request quotas.
Backoff Strategy
Clients should implement exponential backoff with jitter and honor Retry-After on 429.

Sources