Qwen 2.5 VL 7B: benchmark results

Provider: Alibaba. Released 2025-01-26. Access: Open.

Unified ELO 1535 ± 8, rank #599 of 1537 rated models, from 724 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Seeing Culture - Grounding mIoU47.2Overall mIoU (%)100
OpenVLM SEEDBench IMG - Visual Reasoning84Accuracy (%)99.6
OpenVLM MMT-Bench - Spot The Similarity100Score (%)99.3
OpenVLM Q-Bench - Yes-or-No (In-context Distortion)82.9Accuracy (%)99.1
OpenVLM MMT-Bench - Film and Television Recognition100Score (%)99
OpenVLM MMT-Bench - Lesion Grading90Score (%)99
OpenVLM MTVQA - Korean38Accuracy (%)98.7
OpenVLM MTVQA - Vietnamese44.8Accuracy (%)98.7
OpenVLM MMT-Bench - Industrial Produce Anomaly Detection90Score (%)98.3
OpenVLM MMT-Bench - DocVQA90Score (%)98.1
OpenVLM MMT-Bench - Salient Object Detection RGBD65Score (%)97.6
OpenVLM MMT-Bench - Face Retrieval100Score (%)97.3

Interactive version: theaggregate.ai/model?slug=qwen-2-5-vl-7b · How It Works · Data refreshed daily, snapshot 2026-09-25.