WeVisOne-VL - Referring Ref-L4: leaderboard
WeVisOne-VL official release:greedy decoding,temperature0,max1960visual tokens/image;Qwen3-VL Instruct and Qwen3.5 Thinking baselines. Generative visual grounding/perception,not image generation. Classical detectors and specialist OCR systems are outside the VLM cohort;absent-target rejection scores require no box;other columns are meanF1 acrossIoU thresholds0.5–0.95. Score (%).
Metric: Score (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 4B Instruct | 72.5 | #512 |
| 2 | Qwen 3 VL 8B Instruct | 72.4 | #418 |
| 3 | Qwen 3.5 9B (Thinking) | 72 | #388 (Qwen 3.5 9B) |
| 4 | Qwen 3.5 4B (Thinking) | 70.4 | #493 (Qwen 3.5 4B) |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=wevisone-vl-referring-ref-l4 · How It Works · Data refreshed daily, snapshot 2026-10-24.