Qwen 3.5 27B — benchmark results

Alibaba's dense open Qwen 3.5 27B multimodal model. Provider: Alibaba. Released 2026-02-24. Access: Open.

Unified ELO 1670 ± 13, rank #321 of 1776 rated models, from 207 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EgoCoT-Bench71.28Mean (self-reported)100
LLM Stats (Hallusion Bench)70Score (%)100
SVFSearch95.4Overall Acc. (self-reported)100
VANTAGE-Bench - Spatial79.04Spatial (%)100
AI for Education Pedagogy - Technology88.68Accuracy (%)98.9
AI for Education Pedagogy - Science93.99Accuracy (%)98
FMNB Leaderboard100Score97.9
StemBind41.3F Overall (self-reported)95.7
MATH-MC Level 399.37Accuracy (%)95.6
LLM Stats (MathVista-Mini)87.8Score (%)95.5
K-MetBench83Accuracy (self-reported)94.8
MATH-MC Level 299.09Accuracy (%)94.1

Interactive version: theaggregate.ai/model?slug=qwen-3-5-27b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.