Physical AI Bench Generation

Physical AI Bench generation leaderboard for world models predicting future states across autonomous driving, robotics, industry, human, physics, and common-sense scenarios.

17rows
overallprimary metric
2026-05-06sampled

Metadata

Metrics

Overall, Domain, Quality, Common Sense, AV, Robot, Industry, Human, Physics, Subject Consistency, Background Consistency, Motion Smoothness, Aesthetic Quality, Imaging Quality, Overall Consistency, I2V Subject, I2V Background

Latest Results

Rank Subject Overall Model Match Provenance Sampled
1 Source 82.60 Imported 2026-05-06
2 Veo-3 82.10 Imported 2026-05-06
3 Cosmos-Predict2.5-14B 81 Imported 2026-05-06
4 Cosmos-Predict2.5-2B 81 Imported 2026-05-06
5 Wan2.2-I2V-A14B 80.60 Imported 2026-05-06
6 Wan2.2-TI2V-5B 80.40 Imported 2026-05-06
7 Cosmos-Predict2-14B-Video2World 80 Imported 2026-05-06
8 Wan2.1-I2V-14B-720P 79.80 Imported 2026-05-06
9 Cosmos-Predict2-2B-Video2World 79.60 Imported 2026-05-06
10 MAGI-1-24B 78.50 Imported 2026-05-06
11 CogVideoX1.5-5B-I2V 78.30 Imported 2026-05-06
12 CogVideoX-5b-I2V 77.90 Imported 2026-05-06
13 LTX-Video-13B 77.90 Imported 2026-05-06
14 HunyuanVideo-I2V 77.40 Imported 2026-05-06
15 LTX-Video-2B 76.90 Imported 2026-05-06
16 MAGI-1-4.5B 76.90 Imported 2026-05-06
17 DynamiCrafter_1024 69.70 Imported 2026-05-06