gpt-sw3-356m-instruct — benchmark results

Provider: AI Sweden. Released 2023-04-28. Access: Open.

Unified ELO 1051 ± 37, rank #1768 of 1776 rated models, from 94 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval Icelandic NLU - NQII33.29Reading comprehension Score (%)34.7
EuroEval Icelandic Common Sense Reasoning1.39Common Sense Reasoning Average Score (%)26.9
Icelandic LLM - WinoGrande-IS56.53Score (%)24.2
EuroEval Norwegian NLU - Norquad22.37Reading comprehension Score (%)20.4
EuroEval Faroese NLU - FONE36.75Named entity recognition Score (%)19.5
EuroEval Faroese NLU - FoQA6.96Reading comprehension Score (%)18.4
EuroEval Icelandic NLU13.61NLU Average Score (%)18.1
EuroEval Swedish NLU - Multi Wiki QA SV17.03Reading comprehension Score (%)17.8
EuroEval Icelandic NLU - ScaLA IS0Linguistic acceptability Score (%)17.4
EuroEval Icelandic NLU - MIM-GOLD NER16.61Named entity recognition Score (%)16.9
EuroEval Icelandic NLU - Hotter and Colder Sentiment4.54Sentiment classification Score (%)16.8
EuroEval Norwegian8.14Average Score (%)16

Interactive version: theaggregate.ai/model?slug=gpt-sw3-356m-instruct · How the rankings work · Data refreshed daily, snapshot 2026-07-22.