STEM & Reasoning Benchmark

AIME 2026 (DE)

Aleph Alpha's official German translation of AIME 2026, evaluated with the AA eval-framework (pass@1, no tools; same setup as EN AIME 2026 on the Kolibri card).

Category

STEM & Reasoning

Score type

percent

Scale

100%

higher is better

Models tested

1

What is AIME 2026 (DE)?

AIME 2025 uses the problems of the American Invitational Mathematics Examination 2025 (I & II) as a benchmark. These are competition math problems that only have their exact numeric answer graded. Because the problems are new and public, a model can barely rely on training-data memorisation — the year variants show whether a model really computes or just remembers.

What does it test?

AIME 2026 (DE) measures hard reasoning in science, technology, engineering and math. In detail, it checks:

  • competition-level math problem solving
  • exact numeric answers (graded answers must be exact)
  • multi-step reasoning without tools (unless explicitly 'with tools')
  • fresh problems that defeat memorised answer lists

How to read the score

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

Frequently asked questions

What is the AIME 2026 (DE) benchmark?

AIME 2025 uses the problems of the American Invitational Mathematics Examination 2025 (I & II) as a benchmark. These are competition math problems that only have their exact numeric answer graded. Because the problems are new and public, a model can barely rely on training-data memorisation — the year variants show whether a model really computes or just remembers.

What does AIME 2026 (DE) test?

AIME 2026 (DE) mainly measures hard reasoning in science, technology, engineering and math. Concretely, it checks: competition-level math problem solving; exact numeric answers (graded answers must be exact); multi-step reasoning without tools (unless explicitly 'with tools').

How is the AIME 2026 (DE) score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get AIME 2026 (DE) data and scores?

The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/aime-2026-de/scores) or pull it directly from the Open LLM Radar API with any HTTP client.

Model scores 1 model tested

Reset
# Model Score Date Source Verified
1 Kolibri-1
Aleph Alpha
90%
03.10.2026 Report Verified

What now?

You've seen how models score on AIME 2026 (DE) — here's what to do next.