FlagEval EmbodiedVerse - SAT Leaderboard — The Aggregate

FlagEval EmbodiedVerse - SAT: leaderboard

Metric: Score. Source: flageval.baai.ac.cn. 24 models tracked.

Top models

#	Model	Score
1	GPT-5	86
2	Qwen 3 VL 235B A22B Instruct	79.33
3	GPT-4.1	78.67
4	Qwen 3 VL 30B A3B (Thinking)	78
5	Claude Sonnet 4.5	77.33
6	Qwen 3 VL 235B A22B (Thinking)	77.33
7	Gemini 2.5 Pro	76
8	Qwen 3 VL 30B A3B Instruct	72
9	Qwen 2.5 VL 32B Instruct	70.67
10	Mistral Small 3.2	67.33
11	InternVL3-78B	57.33
12	O4 Mini	55.33
13	InternVL3-8B	54.67
14	Qwen 2.5 VL 7B Instruct	43.33
15	Phi-4 Multimodal Instruct	41.33

Interactive version: theaggregate.ai/benchmark?slug=flageval-embodiedverse-sat · How It Works · Data refreshed daily, snapshot 2026-09-05.