Aleph Alpha's sovereign open-weight Kolibri scores 96.9% on AIME, beating 3x bigger MoE models
TejasKumar_ · x · 2026-10-03
Aleph Alpha released Kolibri, a sovereign open-weight German model (78B total params). Each layer has 384 tiny specialists with a router activating 6 per word, so it effectively runs 3.5B params per token — and scores 96.9% on AIME, beating every MoE model tested including ones 3x bigger; only a dense model doing 8x the work wins. It's self-hostable. Its tokenizer needs 15% fewer tokens than GPT-5's on the German constitution ("bundesverfassungsgericht": 6 tokens vs 2), making German cheaper, faster, and more context-efficient; in their evals it tops every open model its size in both English and German.
More from Models
- Observer: Google and Anthropic are winning the public AI race, OpenAI's lead now in doubt — imjustnewatai · 2026-11-19
- Three Google flagship models in eight days fuel speculation about hidden RSI work — haider1 · 2026-11-19
- Mystery Model 'Gestalt' Tops New Frontier Design Benchmark by 214 Points — soleio · 2026-11-19
- mrfakename releases abliterated GLM-5.3-Flash and Qwen3.8-Flash-Next — mrfakename0 · 2026-11-19
- Physician: AI competition now measured in weeks, GPT-6 versions 7 days apart — ramos_casals · 2026-11-19
- micro1's flow qc model screens training data for benchmark leaks before release — Exp_Mark · 2026-11-19