Qwen 3 1.7B: benchmark results

Alibaba's Apache-2.0 1.7B dense Qwen3 model (April 2025) with switchable thinking and non-thinking modes, roughly matching Qwen2.5-3B. Provider: Alibaba. Released 2025-04-28. Access: Open.

Unified ELO 1419 ± 1, rank #1067 of 1392 rated models, from 214 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
MERA Code - ruHumanEval66.59pass@1 (%)87.7
MERA - LCS20.4Accuracy (%)81.3
MERA - ruMultiAr43.85EM (%)76.1
MERA Code - ruCodeEval50.24pass@1 (%)75.4
MERA Code - UnitTests24.89CodeBLEU (%)72.3
EuroEval Finnish NLU - Tydiqa FI64.78Reading comprehension Score (%)66.1
EuroEval Faroese NLU - ScaLA FO6.23Linguistic acceptability Score (%)62
EuroEval Portuguese Knowledge49.5Knowledge Average Score (%)61
EuroEval Spanish NLU - Sentiment Headlines ES42.57Sentiment classification Score (%)60.6
MERA Code0.32Total score60
EuroEval German NLU - Germanquad53.5Reading comprehension Score (%)59.4
EuroEval Spanish Knowledge48.52Knowledge Average Score (%)58.6

Interactive version: theaggregate.ai/model?slug=qwen-3-1-7b · How It Works · Data refreshed daily, snapshot 2026-09-05.