gpt-sw3-356m: benchmark results

Provider: AI Sweden. Access: Open.

Unified ELO 1197 ± 20, rank #2903 of 2928 rated models, from 97 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval Icelandic NLU - ScaLA IS1.57Linguistic acceptability Score (%)40.3
EuroEval Icelandic Common Sense Reasoning2.56Common Sense Reasoning Average Score (%)34.1
EuroEval Icelandic NLU - NQII32.08Reading comprehension Score (%)32.2
EuroEval Swedish NLU - Multi Wiki QA SV36.94Reading comprehension Score (%)28.9
EuroEval Norwegian NLU - Norquad34.55Reading comprehension Score (%)28
EuroEval Danish NLU - Multi Wiki QA DA35.67Reading comprehension Score (%)27.9
EuroEval English NLU - SST-553.65Sentiment classification Score (%)22.5
EuroEval English NLU - SQuAD51.18Reading comprehension Score (%)20.9
Open LLM Leaderboard v1 - TruthfulQA MC242.55MC2 (%) (0-shot)20.8
EuroEval Faroese NLU - FoQA9.43Reading comprehension Score (%)19.9
EuroEval Norwegian11.77Average Score (%)19.7
EuroEval Icelandic NLU - MIM-GOLD NER18.47Named entity recognition Score (%)19.4

Interactive version: theaggregate.ai/model?slug=gpt-sw3-356m · How It Works · Data refreshed daily, snapshot 2026-09-23.