Kosmos2: benchmark results

Provider: Microsoft. Released 2023-06-29. Access: Open.

Aggregate Rating 1154 ± 49, rank #9796 of 10552 rated models, from 454 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
OpenVLM MMT-Bench - Gaze Estimation35Score (%)93
CoreCognition Image + Video Conservation66.97Accuracy (%)91.6
OpenVLM MMT-Bench - Font Recognition35Score (%)88.1
OpenVLM MMT-Bench - Behavior Anomaly Detection60Score (%)84.5
OpenVLM MMT-Bench - Color Constancy57.1Score (%)82.3
OpenVLM MMT-Bench - Animal Keypoint Detection40Score (%)82
OpenVLM COCO Captions14.5CIDEr81.9
OpenVLM MMT-Bench - Jigsaw Puzzle Solving35Score (%)81.6
OpenVLM MMT-Bench - Handwritten Retrieval35Score (%)81.1
OpenVLM MMT-Bench - I2IT37.5Score (%)78.2
OpenVLM MMT-Bench - Image Colorization40Score (%)73.1
OpenVLM MMT-Bench - 3D Indoor Recognition35Score (%)68.2

Interactive version: theaggregate.ai/model?slug=kosmos2 · How It Works · Data refreshed daily, snapshot 2026-11-16.