ERQA: leaderboard
Embodied Reasoning Question Answering benchmark with visual questions about space, trajectories, actions, state estimation, and multi-view reasoning.
Metric: Score (%). Source: llm-stats.com. Status: years away from saturation. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.8 Max | 77.8 |
| 2 | Qwen3.8-Flash-Next | 72.3 |
| 3 | Qwen 3.8 Flash | 72.3 |
| 4 | Seed 2.1 Pro | 72 |
| 5 | Seed 2.1 Turbo | 71.3 |
| 6 | Qwen 3.7 Plus | 69.8 |
| 7 | GPT-5 | 65.7 |
| 8 | Qwen 3.6 Plus | 65.7 |
| 9 | Qwen 3.8 27B | 65.5 |
| 10 | Qwen 3.5 35B A3B | 64.8 |
| 11 | Muse Spark | 64.7 |
| 12 | O3 | 64 |
| 13 | Qwen 3.6 27B | 62.5 |
| 14 | Qwen 3.5 122B A10B | 62 |
| 15 | Qwen 3.5 27B | 60.5 |
Interactive version: theaggregate.ai/benchmark?slug=erqa · How It Works · Data refreshed daily, snapshot 2026-09-05.