Qwen 3.5 35B A3B Base: benchmark results

Provider: Alibaba. Released 2026-02-24. Access: Open.

Unified ELO 1684 ± 22, rank #264 of 1516 rated models, from 312 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
QIMMA - 3LM STEM95.33Accuracy (%, log-likelihood multiple choice)100
Yandex AliceAI Cards - Codeforces C++ (pass@8)73.7Pass@8 (%)100
Yandex AliceAI Cards - EduBench Math80Score (%)100
Yandex AliceAI Cards - GPQA47.1Accuracy (%)100
Yandex AliceAI Cards - GSM8K (CoT)90.4Accuracy (%)100
Yandex AliceAI Cards - HumanEval88.3Score (%)100
Yandex AliceAI Cards - MBPP75.4Score (%)100
Yandex AliceAI Cards - MMLU84.4Accuracy (%)100
Yandex AliceAI Cards - RULER 128K90.1Score (%)100
Yandex AliceAI Cards - YExtract42.4Score (%)100
EuroEval French NLU77.13NLU Average Score (%)98.9
EuroEval Italian NLU - ScaLA IT67.65Linguistic acceptability Score (%)98.4

Interactive version: theaggregate.ai/model?slug=qwen-3-5-35b-a3b-base · How It Works · Data refreshed daily, snapshot 2026-09-24.