ReaLM - Clean Accuracy: leaderboard

Metric: Accuracy (%) on the unperturbed images of the 832 ReaLM physical-world cases (driving, manipulation, grasping, physics, egocentric, diagnostics and scene QA; image, question and ground-truth answer), black-box one-shot evaluation at temperature 0, a Qwen3-8B judge extracting the answer; higher is better. Source: arxiv.org. Saturation forecast: Around May 2027. 13 models tracked.

Top models

#ModelScore
1Qwen 3.5 122B A10B72.8
2Qwen 3.6 Flash71.7
3Qwen 3.5 27B71.2
4Qwen 3.5 9B68.9
5Gemini 3 Flash68.5
6GPT-4.1 Mini68.3
7Kimi K2.568.3
8Seed 2.0 Mini67.2
9Qwen 3 VL 32B66.9

Interactive version: theaggregate.ai/benchmark?slug=realm-clean-accuracy · How It Works · Data refreshed daily, snapshot 2026-09-29.