cogvlm2-llama3-chat-19B: benchmark results

Provider: Zhipu.

Aggregate Rating 1419 ± 45, rank #5300 of 10489 rated models, from 19 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
CoreCognition Image only Intuitivephysics68.1Accuracy (%)100
CoreCognition Image only Permanence78.13Accuracy (%)92.8
CoreCognition Image only Perspective43.66Accuracy (%)79.6
CoreCognition Image only Boundary76.31Accuracy (%)77.6
CoreCognition Image only Mean60.08Accuracy (%)76.3
CoreCognition Image only Continuity61.72Accuracy (%)57.9
CoreCognition Image only Perceptualconstancy60.37Accuracy (%)56.6
CapArena-Auto-8.5CapArena-Auto Score (avg)52.4
CapArena-Auto vs CogVLM-19B0CapArena-Auto Score (vs CogVLM-19B)52.4
CapArena-Auto vs MiniCPM-8B31CapArena-Auto Score (vs MiniCPM-8B)52.4
CoreCognition Image only Intentionality65.71Accuracy (%)50
CoreCognition Image only Toolusing92.29Accuracy (%)49.3

Interactive version: theaggregate.ai/model?slug=cogvlm2-llama3-chat-19b · How It Works · Data refreshed daily, snapshot 2026-11-20.