CoreCognition Image only Conservation: leaderboard

CoreCognition Image only Conservation — CoreCognition native component accuracy and reported mean. Image+Video, image-only and single-image populations are separate; no pooling or replacement across cohorts. Metric: Accuracy (%). Source: https://grow-ai-like-a-child.github.io/CoreCognition/leaderboard

Metric: Accuracy (%). Source: grow-ai-like-a-child.github.io. Saturation forecast: Rough model projection: around 2026. 77 models tracked. Updated 16 Nov 2026.

Top models

#ModelScoreOverall rank
1Yi-VL-34B69.21#6990
2cogvlm-grounding-generalist67.94
3ShareGPT4V-13B67.62#8424
4wemm-chat-2k-cn67.62
5visualglm-6B66.98
6fuyu-8B66.98
7instructblip-flan-t5-xl66.67
8step-1o-turbo-vision66.35
9InstructBLIP-Flan-T5-XXL66.03
10InternLM-XComposer2-4KHD65.71#6477
11InstructBLIP-Vicuna-7B65.4
12smolvlm-dpo65.08
13Cambrian-8B64.76#6834
14mgm-7B-hd64.44
15omchat64.44

Interactive version: theaggregate.ai/benchmark?slug=corecognition-image-only-conservation · How It Works · Data refreshed daily, snapshot 2026-11-16.