Embodied Analysis

Unified Evaluation for Embodied Intelligence

Independent and comprehensive evaluation for the perception, generation, action, and physical agentic capabilities of embodied AI models

Evaluation methods
Embodied AnalysisPhysical Agent Capability Index
Evaluation Method
GPT 6 Astra 1362
Gemini 4 Argon 1331
Claude Opus 5.5 1313
Claude Opus 5 1289
Claude Fable 5.1 1182
Muse Wonder 1040
GPT 6 Sol 1012
Claude Opus 4.8 990
MiniMax M3.1 Flash 985
Gemini 3.8 Flash 972
Grok 4.6 972
DeepSeek V4.1 Flash 921
Qwen 3.8 Max 915
GLM 5.3 Flash 903
π0.5 896
Muse Spark 1.3 844
Kimi K3 828
GPT 5.5 823
Inkling 423
19 modelsElo Average
Embodied AnalysisWorld Model Capability Index
Evaluation Method
Wan 3 1161
MiniMax H3 1129
Kling v3 Pro 1101
Seedance 2 1082
Gemini Omni Flash 953
Grok Imagine 1.5 932
Cosmos 3 Super 831
Veo 3.1 812
8 modelsElo Average
Physical Agent

Physical Agent Evaluation

Physical agents are evaluated on their capabilities to drive physical embodiments to complete complex long-horizon tasks.

Read More →
01

Goal Completion

Does the embodiment complete the specified task within time limit?
02

Generalization

Does the model generalize across different tasks and embodiments?
03

Efficiency

Does the model produce a clean and efficient execution trajectory?
World Model

World Model Evaluation

World models are evaluated on multiple axes including physical plausibility, goal completion, and instruction following.

Read More →
01

Physical Plausibility

Do gravity, contact, and object solidity remain plausible, without clipping or teleportation?
02

Goal Completion

Does the clip reach the goal state described by the task?
03

Instruction Following

Does the robot follow the instruction while keeping the same embodiment?