Kolibri: A Sovereign European Model on the Pareto Frontier

Benchmark Model Rank Results
agents-on-bench-telecomKolibri#2Accuracy: 94.7
agents-on-bench-airlineKolibri#4Accuracy: 76.7
mathematical-reasoning-on-aime-2025Kolibri#4Accuracy: 96.9
mathematical-reasoning-on-aime-2026Kolibri#4Accuracy: 96.0
agents-on-bench-retailKolibri#6Accuracy: 69.9
question-answering-on-humanity-s-last-exam-hleKolibri#26Accuracy: 21.5
multiple-choice-question-answering-mcqa-on-gpqa-diamondKolibri#41Accuracy: 84.3
multi-task-language-understanding-on-mmlu-proKolibri#44Accuracy: 80.0
code-generation-on-terminal-bench-2-1Kolibri (Terminus-2)#47Accuracy: 27.7
code-generation-on-swe-bench-verifiedKolibri (OpenCode)#53Accuracy: 66.4
deep-research-agents-on-browsecompKolibri#57Accuracy: 29.4
code-generation-on-livecodebench-v6Kolibri#63Pass@1: 0.859