BLIP-2-OPT-2.7B: benchmark results

Provider: Other.

Aggregate Rating 1400 ± 52, rank #6172 of 10489 rated models, from 23 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
CoreCognition Image only Conservation61.27Accuracy (%)78.3
Epoch AI - Science Qa74.17Score78.3
CoreCognition Image only Perceptualconstancy57.81Accuracy (%)50
CoreCognition Image only Perspective32.82Accuracy (%)38.2
Mind's Eye - Hierarchical Pattern Equivalence22.7Accuracy (%) on the hierarchical pattern equivalence task of35.3
CoreCognition Image only Spatiality41.48Accuracy (%)17.8
Mind's Eye - Symmetric Structures17Accuracy (%) on the symmetric structures task of Mind's Eye 17.6
CoreCognition Image only Permanence21.88Accuracy (%)17.1
CoreCognition Image only Hierarchy30.83Accuracy (%)11.8
Mind's Eye - Dynamic Structural Correspondence18.3Accuracy (%) on the dynamic structural correspondence task o11.8
Mind's Eye - Mental Composition24Accuracy (%) on the mental composition (net to solid) task o11.8
CoreCognition Image only Boundary45.41Accuracy (%)10.5

Interactive version: theaggregate.ai/model?slug=blip-2-opt-2-7b · How It Works · Data refreshed daily, snapshot 2026-11-20.