FlagEval EmbodiedVerse - EgoPlan-Bench2: leaderboard
BAAI FlagEval’s EmbodiedVerse suite, EgoPlan-Bench2 component: egocentric planning, where a first-person video and a goal have to yield the next action. Scored on FlagEval’s common scale so the thirteen component boards read against each other.
Metric: Score. Source: flageval.baai.ac.cn. Saturation forecast: Not forecast. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 30B A3B Instruct | 58.8 | #384 |
| 2 | Qwen 3 VL 235B A22B Instruct | 57.74 | #273 |
| 3 | Qwen 2.5 VL 32B Instruct | 53.37 | #453 |
| 4 | Qwen 2.5 VL 72B Instruct | 52.69 | #377 |
| 5 | GPT-5 | 51.85 | #98 |
| 6 | Gemini 2.5 Pro | 50.15 | #152 |
| 7 | Qwen 3 VL 235B A22B (Thinking) | 49.39 | #238 (Qwen 3 VL 235B A22B) |
| 8 | Qwen 3 VL 30B A3B (Thinking) | 48.03 | #351 (Qwen 3 VL 30B A3B) |
| 9 | InternVL3-78B | 47.46 | #352 |
| 10 | GPT-4.1 | 43.22 | #248 |
| 11 | InternVL3-8B | 39.21 | #620 |
| 12 | Qwen 2.5 VL 7B Instruct | 38.46 | #654 |
| 13 | Claude Sonnet 4.5 | 33.61 | #142 |
| 14 | O4 Mini | 33.46 | #180 |
| 15 | Mistral Small 3.2 | 31.57 | #605 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=flageval-embodiedverse-egoplan-bench2 · How It Works · Data refreshed daily, snapshot 2026-10-20.