Model · Language model

Kolibri 1

Aleph Alpha's sovereign English-German mixture-of-experts reasoning model, 78.1B total parameters with 3.46B active per token, trained from scratch for public administration and regulated industry and published as FP8 weights under Apache 2.0.

Developer

Aleph Alpha Research GmbH

Distribution

Downloadable weights

Artifact licensing

Open-licensed artifacts

License scope

Reviewed release artifacts and mandatory terms; not an assessment of training code or training data openness.

Licenses

Apache License 2.0

Boundary

This record represents the Kolibri 1 release of 3 October 2026, the version whose weights are published as Aleph-Alpha/Kolibri-1 and whose serving parsers are named kolibri1. It excludes Kolibri Origin, the earlier pipeline-validating iteration the developer states had "no public release"; the Pharia enterprise platform and any managed endpoint, since no first-party API is documented for this release; the aleph-alpha-inference serving package, which is a separate repository; and any downstream fine-tune, quantization, or re-upload of these weights.

Open in the directory → Open official model page ↗

Editorial ratings appear in the directory beside the profile they belong to and are never compared across collections.