System status
All systems operational
The governed gateway that Alexandria servers reach frontier models through, and the model tiers it routes to. Open-weight models run in your own perimeter and are listed separately, below.
- Inference gateway
- Responding.
- Authentication
- Keys validating.
- Claude Opus 4.8
- Anthropic
- Claude Sonnet 5
- Anthropic
- Claude Haiku 4.5
- Anthropic
Last updated
Not yet routed
Published so the roadmap is legible, not because the gateway serves them today. Until a tier appears in the table above, no request reaches it.
- Claude Fable 5
- Anthropic · not yet routed
- GPT 5.6
- OpenAI · not yet routed
In your perimeter
Open-weight models served on the server’s own GPUs, inside a Keystone environment. They do not pass through this gateway, so nothing here reports on them. A prompt answered by one never leaves your building.
- Kimi K3
- Open weight · served on-server
- GLM 5.2
- Open weight · served on-server
Where prompts can go
Work from an Alexandria server reaches exactly one of three places: Keystone, Anthropic, OpenAI. Nowhere further. Open-weight models stay on your hardware; frontier work is relayed by Keystone’s gateway to the model provider and to no one else. Anthropic is the only provider routed today. The full detail is in the Privacy notice.
Something wrong that isn’t reflected here? hello@keystone.computer