gpt-sw3-6.7B-v2-instruct: benchmark results

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

Unified ELO 1280 ± 21, rank #2754 of 2928 rated models, from 36 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval Norwegian NLU - Norec34.39Sentiment classification Score (%)35.6
EuroEval Danish NLU - ScaLA DA10.99Linguistic acceptability Score (%)33.9
EuroEval Norwegian Knowledge - Idioms NO0.75MCC (x100)33.6
EuroEval Swedish NLU - ScaLA SV10.92Linguistic acceptability Score (%)32
EuroEval Norwegian NLU - ScaLA NN5.11Linguistic acceptability Score (%)31.6
EuroEval Icelandic Knowledge3.28Knowledge Average Score (%)31.3
EuroEval Danish Common Sense Reasoning11.08Common Sense Reasoning Average Score (%)30.1
EuroEval Swedish NLU - Multi Wiki QA SV38.22Reading comprehension Score (%)29.2
EuroEval Swedish Common Sense Reasoning10.9Common Sense Reasoning Average Score (%)29
EuroEval Norwegian Knowledge4.46Knowledge Average Score (%)26.6
Open LLM Leaderboard v1 - GSM8K6.37Accuracy (%) (5-shot)25.5
EuroEval Norwegian Knowledge - NRK Quiz QA8.16MCC (x100)25.2

Interactive version: theaggregate.ai/model?slug=gpt-sw3-6-7b-v2-instruct · How It Works · Data refreshed daily, snapshot 2026-09-23.