Qwen3.8 Launch - RefAdv-S: leaderboard
Ref-Adv-S is the public subset of Ref-Adv, a referring-expression comprehension benchmark that removes shortcuts with long expressions, negations and several hard distractors per image, so a model must reason to ground the described object. Scores come from the multimodal table of the Qwen3.8-Max launch post; percent, higher is better.
Metric: Score (%). Source: qwen.ai. Saturation forecast: Not forecast. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.8 Max | 80.2 | #34 |
| 2 | Qwen 3.7 Plus | 73 | #98 |
| 3 | Gemini 3.1 Pro (Preview) | 71.9 | #64 |
| 4 | GPT-5.6 Sol | 69.2 | #22 |
| 5 | Claude Fable 5 | 68.6 | #20 |
| 6 | Claude Opus 4.8 | 61.7 | #39 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=qwen3-8-launch-refadv-s · How It Works · Data refreshed daily, snapshot 2026-11-05.