Qwen 2.5 VL 7B Instruct: benchmark results

Alibaba Qwen 2.5 VL 7B vision-language instruction-tuned checkpoint. Provider: Alibaba. Released 2025-01-28. Access: Open.

Unified ELO 1450 ± 1, rank #941 of 1392 rated models, from 84 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
LLM Stats (DocVQA)95.7Score (%)96.3
LLM Stats (TextVQA)84.9Score (%)93.3
LLM Stats (InfoVQA)82.6Score (%)88.9
EVALITA - MAIA-MC81.77CPS83.3
KOFFVQA - Object Attributes78.33Score (%)77.2
KOFFVQA - Korean OCR85Score (%)75.3
WildRoadBench23.6AP50 (self-reported)75
KOFFVQA - Document Understanding74.67Score (%)69.1
LLM Stats (ChartQA)87.3Score (%)68
KOFFVQA - Hallucination and Robustness75Score (%)67.9
EVALITA - text-entailment78.23CPS66
KOFFVQA - Table Understanding61.67Score (%)63

Interactive version: theaggregate.ai/model?slug=qwen-2-5-vl-7b-instruct · How It Works · Data refreshed daily, snapshot 2026-09-05.