AI Potluck
Model components / Evaluation code

EuroEval

EuroEval

Benchmarking framework for language models across 30+ European languages, maintained at the Alexandra Institute and formerly known as ScandEval. It evaluates encoders, decoders and instruction-tuned models alike and publishes per-language leaderboards at euroeval.com.

Formerly ScandEval; the legacy scandeval package still draws about 2.2K downloads a month, and even summed with euroeval's the product stays under 10K, so the rename does not move the band. Verified 2026-09-01 via the EuroEval/EuroEval repository, its license endpoint, its README, euroeval.com, the PyPI JSON backlink and pypistats.

Openness

5 high confidence
5.0
license
MIT(OSI)
source
public(GitHub)
core-gated
ungated

MIT and fully public: the LICENSE body is the stock MIT text (copyright Dan Saattrup Smart and Alexandra Instituttet A/S) with no appended terms. The package installs from PyPI, dataset-creation scripts ship in the repo, and no pricing, paid tier or license key appears in the README or on the docs site.

Adoption

1 high confidence
1.0

The euroeval package draws 750 PyPI downloads in the trailing 30 days, banding at level 1 (<10K). The package is the documented install path and is backlink-verified. The legacy scandeval package (the project's pre-rename name) still draws 2,207 a month; summed the product stays well under 10K, so the band holds across the rename.

Capability

3 medium confidence
3.0

The deepest European-language coverage in the category — 30+ languages, all model classes, reproducible datasets and public leaderboards — but a single-domain harness with a narrow task surface and one primary maintainer, an order below lighteval's 1000+ tasks and backend spread, so one below lighteval beside bigcodebench and olmes.

  • https://euroeval.com recorded 2026-09-01

    "the standard benchmark for evaluating language models across 30+ European languages"; supports encoders, decoders, encoder-decoders, base and instruction-tuned models; publisher Alexandra Institute; interactive leaderboards

  • https://raw.githubusercontent.com/EuroEval/EuroEval/main/README.md recorded 2026-09-01

    dataset reproduction scripts per language; formerly ScandEval; maintainer and docs links

Verified 2026-09-01