XTC-Bench - Object Retrieval: leaderboard

XTC-Bench (2026) object-retrieval questions on COCO and Visual Genome images: the model names the object that has the given attributes, judged against the scene-graph reference. Score 0-1, shown times 100; higher is better.

Metric: Understanding score (0-1, times 100) on object-retrieval questions (which object has the given attributes), a Qwen3-235B judge scores each fact 0-5, normalized to 0-1 and shown times 100, over XTC-Bench's 2,000 scene-graph-annotated images from COCO 2017 val and Visual Genome (over 31,000 atomic facts about objects, attributes and relations); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-571.8#185
2Gemini 2.5 Flash Image68.6
3BLIP3-o8B67.9
4BAGEL-7B-MoT67.4#4668
5OmniGen267.2#4625
6Janus-Pro-7B61.6#6889
7Show-o2-7B58.3
8Tar-7B55.7
9Show-o41.5
10MMaDA-8B37.7

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=xtc-bench-object-retrieval · How It Works · Data refreshed daily, snapshot 2026-11-22.