MECoBench - Parallel (2 Agents): leaderboard
MECoBench (2026) studies multimodal agent collaboration on 96 VirtualHome household rearrangement tasks, where agents act from first-person views and exchange one broadcast message each before acting. This board is the success rate (%) of two agents sharing all rooms and a 60-step budget; higher is better.
Metric: Success rate (%) in the parallel setting (every agent may enter every room; 60 total steps shared by the team) with two agents, on 96 VirtualHome household rearrangement tasks (2 to 7 subgoals each) in which multimodal agents act from four-view first-person observations; agents exchange one broadcast message each before acting; printed 0-1 and x100 here; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 9B | 93.8 | #803 |
| 2 | Qwen 3.5 27B | 93.8 | #359 |
| 3 | GPT-5.4 | 89.6 | #123 |
| 4 | Gemma 4 31B (IT) | 83.3 | #342 |
| 5 | Gemma 4 26B A4B (IT) | 81.2 | #549 |
| 6 | Qwen 3 VL 32B Instruct | 80.2 | #585 |
| 7 | Gemini 3.1 Pro (Preview) | 77.1 | #94 |
| 8 | GLM-4.6V | 71.9 | #701 |
| 9 | Qwen 3 VL 235B A22B Instruct | 66.7 | #522 |
| 10 | Qwen 3 VL 8B Instruct | 45.8 | #981 |
| 11 | Llama 4 Scout Instruct | 25 | #1543 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mecobench-parallel-2-agents · How It Works · Data refreshed daily, snapshot 2026-11-13.