A single OpenAI-compatible endpoint at https://api.ebotservers.com/v1. Point any OpenAI SDK, Cursor, Cline, Claude Code or DeepSeek Harness at it with your key and every model in this catalog is reachable by name - no per-provider setup, no separate billing.

Metered models draw down your monthly allowance, which resets every 30 days on all billing cycles. The Qwen 3.8 27B builder lane runs on dedicated hardware and never touches the allowance - run background jobs, linting and cron work on it freely.

Live telemetry from status.ebotservers.com, the same feed our own monitoring uses. Each model page shows its own 6h / 12h / 24h / 7d history with uptime, time-to-first-byte and per-sample throughput. When a route has no telemetry we show a dash rather than an estimate.

They are announced but not yet verified on our infrastructure. They appear in the catalog so you can see what is queued, but they are not callable until the status changes to live - and they are never counted toward the live model total.

No. We publish a score only when a credible source has released one. A dash means the figure is not published; a dotted value means the source reported it as approximate. We do not extrapolate or estimate benchmark scores.