gpt-sw3-20B-instruct — benchmark results

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

Unified ELO 1313 ± 35, rank #1505 of 1776 rated models, from 54 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval Icelandic NLU - NQII53.88Reading comprehension Score (%)81.5
EuroEval English NLU - SQuAD83.08Reading comprehension Score (%)74.2
EuroEval Norwegian NLU - Norquad67.93Reading comprehension Score (%)64.4
EuroEval English NLU - SST-565.58Sentiment classification Score (%)56.9
EuroEval Icelandic25.68Average Score (%)45.7
EuroEval Icelandic NLU - ScaLA IS1.95Linguistic acceptability Score (%)44.1
EuroEval Norwegian NLU - Norec40.89Sentiment classification Score (%)43.1
EuroEval Norwegian NLU - ScaLA NN8.32Linguistic acceptability Score (%)37.6
EuroEval Icelandic NLU25.68NLU Average Score (%)35.8
EuroEval Swedish NLU - Swerec68.23Sentiment classification Score (%)33.5
EuroEval Swedish NLU - ScaLA SV12.39Linguistic acceptability Score (%)33
EuroEval Norwegian NLU - ScaLA NB9.45Linguistic acceptability Score (%)32.1

Interactive version: theaggregate.ai/model?slug=gpt-sw3-20b-instruct · How the rankings work · Data refreshed daily, snapshot 2026-07-22.