IDEAL-Bench - Object Recognition Recall: leaderboard

Metric: Recognition rate (%): share of in-view ground-truth objects in non-grid rooms that a same-category prediction matches, IDEAL-Bench 3D layout task; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.

Top models

#ModelScore
1Gemini 2.5 Pro89.9
2Claude Sonnet 4.685.4
3GPT-5.482.2
4Gemma 4 31B (IT)81.8
5Qwen 3 VL 30B A3B Instruct79.9
6GLM-4.6V76.8
7Qwen 3 VL 235B A22B Instruct76.7
8Qwen 3 VL 8B Instruct73
9GPT-4o72.8
10InternVL3.5-8B72
11Qwen 2.5 VL 72B Instruct70.6

Interactive version: theaggregate.ai/benchmark?slug=ideal-bench-object-recognition-recall · How It Works · Data refreshed daily, snapshot 2026-09-29.