Models · Coming soon
Frontier open models — coming soon
One OpenAI-compatible endpoint, a curated catalog of the strongest open models — served on
EU-sovereign infrastructure, never trained on your prompts. The launch line-up below is being
provisioned now. Point your SDK, curl, CI or opencode at a single base URL and
switch models by name.
Launch line-up
Available at launch. Exact context windows, per-token pricing and rate limits are published as each model is provisioned.
GLM 5.2
Zhipu AI's flagship general + coding model. Strong tool-use and long-context reasoning.
Coming soon
MiniMax M3
MiniMax's latest high-throughput model — built for fast, cost-efficient agentic workloads.
Coming soon
DeepSeek V4
DeepSeek's frontier reasoning + coding model. A strong default for hard technical tasks.
Coming soon
Kimi 2.7
Moonshot's Kimi — long-context specialist for large documents and codebases.
Coming soon
How it works
- Drop-in OpenAI-compatible — change the base URL, keep your SDK. Select a model by name.
- EU-sovereign — inference on EU infrastructure; control plane on Deutsche Telekom T Cloud in Germany.
- Metadata-only — prompts are never stored or trained on.
- Prepaid credits — transparent per-token pricing, per-key/team budgets and spend alerts.
The catalog will grow beyond the launch line-up. Models are referenced by name, so moving a model between backends is invisible to your code.