InternVL2.5-78B-CoT — benchmark results

Provider: Shanghai AI Lab. Released 2024-12-05. Access: Open.

Unified ELO 1558 ± 11, rank #622 of 1841 rated models, from 69 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open LMM Reasoning - MathVerse - Expression34.4Accuracy (%)96.4
Open LMM Reasoning - DynaMath - Subject-plane geometry39Accuracy (%)93.4
OpenVLM MathVista - TQA74.7Accuracy (%)92.9
OpenVLM MathVista - ARI68.8Accuracy (%)90.8
OpenVLM MathVista - STA84.1Accuracy (%)90.5
OpenVLM MathVista - SCI71.3Accuracy (%)89.8
Open LMM Reasoning - MathVision - counting29.9Accuracy (%)89.3
Open LMM Reasoning - DynaMath - Subject-puzzle test23.5Accuracy (%)88.8
OpenVLM MathVista - VQA58.7Accuracy (%)87.9
Open LMM Reasoning - DynaMath - Subject-plane geometry; Avg53.8Accuracy (%)86.7
OpenVLM MathVista - FQA72.5Accuracy (%)86.7
Open LMM Reasoning - DynaMath - Subject-puzzle test; Avg48.8Accuracy (%)86.2

Interactive version: theaggregate.ai/model?slug=internvl2-5-78b-cot · How It Works · Data refreshed daily, snapshot 2026-07-25.