Subscribe

The trick: Scope Swap

Kolibri's weights really are downloadable under Apache 2.0.

The sovereign label sits beside a signed deal for Aleph Alpha to operate globally as Cohere.

Issue 344 October 202611 receipts3 min

Aleph Alpha calls Kolibri a sovereign open-weight model that matches models with up to four times its active parameters.

Before you read on. Your call?

The weights are on Hugging Face under Apache 2.0, as stated. The benchmark line holds against the rivals shown, but on Aleph Alpha's own harness a dense Qwen3.8 27B scores 79.9 in German against Kolibri's 70.8.

The twist

Two weeks before launch, Aleph Alpha signed a business combination agreement to operate globally as Cohere, still awaiting regulatory approval.

79.9Qwen3.8 27B German score on Aleph Alpha's own harness
70.8Kolibri German score on the same harness
78GB of memory the model card says the FP8 weights need

There’s more to this story.

Membership opens the full investigation, the strongest counterargument and what to do with what you’ve learned.

Start your free month →

First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge. Already a member? Sign in

The trick has a name

We call it Scope Swap: applies to far less than it sounds like. You'll see it again. Learn to spot it →

Say this in tomorrow's meeting“Kolibri's weights really are open under Apache 2.0. But on Aleph Alpha's own German score a dense Qwen model leads by 9.1 points, and the company has signed to operate as Cohere.”

Receipts

  1. Supports aleph-alpha.com: The model can be downloaded with the full weights on Hugging Face and used under the Apache 2.0 license terms.
  2. Supports aleph-alpha.com: Kolibri matches models with up to four times its active parameter count, such as Nemotron 3 Super.
  3. Supports aleph-alpha.com: trained it on infrastructure in Germany and Finland, under European and German law
  4. Context huggingface.co: Model memory footprint: ~78 GB (FP8 weights).
  5. Context huggingface.co: Supporting two languages rather than many is a deliberate choice of depth over breadth.
  6. Context aleph-alpha.com: Customers have full freedom of deployment and intellectual-property safety
  7. Context aleph-alpha.com: today announced the signing of a definitive business combination agreement
  8. Context aleph-alpha.com: Operating globally as Cohere, the unified company will effectively create the first transatlantic sovereign AI solution.
  9. Context aleph-alpha.com: The transaction remains subject to final regulatory approvals.
  10. Refutes tej.as: Qwen3.8 27B scores 80.2 in English and 79.9 in German, but it uses nearly 8 times as many parameters for every token
  11. Context news.ycombinator.com: Qwen3.8 27B beats Kolibri 79.9 vs 70.8 in German in Kolibri's harness on Kolibri's benchmark.

Open the Receipts Pack → What each source proves, every figure traced, and what would change our verdict.

This story is a stable, citable object. If you can falsify a verdict,tell us. Corrections are loud here.