EPIC-Bench - Placement Region: leaderboard
Metric: Manipulation score (0-100) on placement region (where an object can be placed, with a physical feasibility judgement), on EPIC-Bench (6,661 human-annotated image, text and mask tuples from 25 public datasets across 23 fine-grained embodied perception tasks; the model outputs bounding boxes, counts, path points or feasibility judgements that are scored against the masks), zero-shot, averaged over six runs for local models and two for API models; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 89 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 55.25 |
| 2 | Gemini 3 Pro | 53.42 |
| 3 | Gemini 3 Flash (Preview) | 50.89 |
| 4 | GPT-5.5 (Non-reasoning) | 48.7 |
| 5 | Gemini 3 Flash (Preview) (Non-reasoning) | 48.09 |
| 6 | Qwen 3 VL 235B A22B (Thinking) | 46.49 |
| 7 | Claude Sonnet 4.6 (Thinking) | 43.81 |
| 8 | O3 | 43.64 |
| 9 | Gemini 2.5 Pro | 42.78 |
| 10 | Qwen 3.5 397B A17B | 42.51 |
| 11 | Qwen 3.5 Plus | 42.46 |
| 12 | Qwen 3.6 Plus | 41.59 |
| 13 | Seed 1.8 | 40.79 |
| 14 | Qwen 3 VL 30B A3B (Thinking) | 40.72 |
| 15 | Qwen 3 VL 32B (Thinking) | 40.03 |
Interactive version: theaggregate.ai/benchmark?slug=epic-bench-placement-region · How It Works · Data refreshed daily, snapshot 2026-10-07.