GPT-OSS-20B (Low) — benchmark results

GPT-OSS-20B evaluated at the low reasoning-effort setting. Provider: OpenAI. Released 2025-08-05. Access: Open.

Unified ELO 1543 ± 5, rank #643 of 1776 rated models, from 345 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval Dutch Simplification - Duidelijke Taal56.94Score (%)97.7
EuroEval Polish Summarization - PSC25.36Score (%)96.6
EuroEval Spanish NLU - ScaLA ES44.57Linguistic acceptability Score (%)93.4
EuroEval Portuguese NLU - ScaLA PT40.49Linguistic acceptability Score (%)90
EuroEval Slovak NLU - UNER SK63.56Named entity recognition Score (%)89.7
EuroEval Hungarian NLU - Szeged NER73.76Named entity recognition Score (%)89.6
EuroEval Italian NLU - ScaLA IT41.68Linguistic acceptability Score (%)87.4
EuroEval Croatian NLU - WikiANN HR66.58Named entity recognition Score (%)85.7
EuroEval Portuguese NLU58.01NLU Average Score (%)85
EuroEval Spanish NLU53.78NLU Average Score (%)84.9
EuroEval English NLU - ScaLA EN55.84Linguistic acceptability Score (%)84.7
EuroEval Czech NLU - PONER57.61Named entity recognition Score (%)84.6

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