MiniCPM-V-2.6 — benchmark results

OpenBMB's 8B multimodal model (August 2024) built on SigLIP-400M and Qwen2-7B, handling single-image, multi-image and video understanding. Provider: OpenBMB. Released 2024-08-06. Access: Open.

Unified ELO 1398 ± 7, rank #1235 of 1776 rated models, from 998 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
MEGA-Bench Task - MuMA Theory Of Mind Social Goal53.3Task Score (%)100
MEGA-Bench Task - VLN Hindi Next Step46.7Task Score (%)98.8
OpenVLM MMT-Bench - Chart To Text100Score (%)98.8
MEGA-Bench Task - Music Sheet Note Count17.6Task Score (%)97.7
OpenVLM MMT-Bench - Traffic Sign Understanding95Score (%)97.3
MEGA-Bench Task - Egocentric Analysis Single Image77.8Task Score (%)96.5
MEGA-Bench Task - MFC Bench Check Clip Stable Diffusion Generate64.3Task Score (%)96.2
OpenVLM MMT-Bench - Count58.1Score (%)95.9
OpenVLM MMT-Bench - Interactive Segmentation71.4Score (%)95.4
OpenVLM MME - OCR192.5Score95.1
OpenVLM MMT-Bench - Animals Recognition100Score (%)94.9
OpenVLM MMT-Bench - Religious Recognition80Score (%)94.9

Interactive version: theaggregate.ai/model?slug=minicpm-v-2-6 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.