Title: LLM-riktm%C3%A4rken: vems siffror, producerade hur %E2%80%94 LLM EU
Canonical: https://llmeu.com/sv/benchmarks/
Markdown mirror: https://llmeu.com/sv/benchmarks.md
Machine-readable: https://llmeu.com/llms.txt (index) - https://llmeu.com/llms-full.txt (full catalogue) - https://llmeu.com/llms-catalogue.json

# Benchmarks

Ett verifierat p%C3%A5st%C3%A5ende p%C3%A5 LLM EU anger var det kommer fr%C3%A5n: ett dokument du kan %C3%B6ppna, eller ett v%C3%A4rde registrerat p%C3%A5 endpoint-raden.

## [European languages](/sv/benchmarks/eu-languages/)

The EU21 numbers are OpenGPT-X%E2%80%99s, taken from the Teuken 7B Instruct v0.6 model card and the accompanying evaluation preprint; they are machine-translated benchmarks, not native-language test sets. The MMMLU number is Google%E2%80%99s own. LLM EU has run no evaluation and has not verified either harness.

## [Code and software agents](/sv/benchmarks/code/)

Every score in this set is republished from the model publisher%E2%80%99s own model card or vendor report and names that source per entry. LLM EU has run no evaluation, has not reproduced any score, and has not verified the harness settings. Vendor-reported numbers are self-reported and are labelled as such on the model cards. SWE-bench results depend heavily on the agent scaffold; the NVIDIA entry names its scaffold explicitly, and scores from different scaffolds must not be compared directly.

## [Reasoning and mathematics](/sv/benchmarks/reasoning/)

Every score in this set is republished from the model publisher%E2%80%99s own model card or vendor report and names that source per entry. LLM EU has run no evaluation, has not reproduced any score, and has not verified the harness settings. Vendor-reported numbers are self-reported and are labelled as such on the model cards. AIME and GPQA scores are sampling-sensitive: number of samples, temperature and tool access are stated on the individual model cards and are not uniform across these rows.

## [Long context](/sv/benchmarks/long-context/)

Every score in this set is republished from the model publisher%E2%80%99s own model card or vendor report and names that source per entry. LLM EU has run no evaluation, has not reproduced any score, and has not verified the harness settings. Vendor-reported numbers are self-reported and are labelled as such on the model cards. RULER scores come from NVIDIA%E2%80%99s own NeMo Evaluator harness and are not comparable with scores produced by other harnesses. The context-window rows are configuration or vendor claims and say nothing about answer quality at that length.

## [Small models, cost bracket](/sv/benchmarks/cost-quality/)

Every score in this set is republished from the model publisher%E2%80%99s own model card or vendor report and names that source per entry. LLM EU has run no evaluation, has not reproduced any score, and has not verified the harness settings. Vendor-reported numbers are self-reported and are labelled as such on the model cards. LLM EU publishes no price data in this set and has run no cost or throughput benchmark, so the rows must not be read as value for money: each metric is stated per row and the models differ in size, licence and serving requirements.

LLM EU Verified betyder dokumenterad, inte juridiskt certifierad. Det %C3%A4r inte en bed%C3%B6mning av %C3%B6verensst%C3%A4mmelse enligt AI-f%C3%B6rordningen och LLM EU %C3%A4r inte ett EU-organ.