Status

All systems operational

Checked . Every reload runs a fresh health check against each GPU pool.

Answering requestsRegion: Switzerland · liveCheck again

Components

Each GPU pool is probed live when this page loads.

API gateway (Cloudflare edge)Global edgeOperational:Serving
Primary GPU poolSwitzerlandOperational:Ready · 413 ms probe

Routing mode: local. Requests fail over between GPU pools automatically; a pool being down only reduces redundancy.

When a pool has a bad minute

What your agent sees, and what it costs.

Failover between pools

  1. Your request
  2. Primary GPU pool
  3. If busy or down: next pool

The edge sends each check to the primary GPU pool in Switzerland. When more than one pool is configured and the primary is saturated or unhealthy, the gateway retries on the next one. The pools listed above are the ones configured right now, with their regions.

If no pool can answer

The API returns 503 with a retry hint. Nothing is charged and no free request is used. Decide in advance what your agent does meanwhile: for irreversible actions, pausing and asking a human is usually the safer default.

Error handling in the docs