Qwen3.8 Launch - VLMsAreBiased: leaderboard
VLMs Are Biased tests whether vision-language models read what is actually in an image or answer from prior knowledge, using counting questions on counterfactually edited images of familiar subjects, such as a logo with an extra stripe. Scores come from the multimodal table of the Qwen3.8-Max launch post; percent, higher is better.
Metric: Score (%). Source: qwen.ai. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.8 Max | 88.3 | #34 |
| 2 | Gemini 3.1 Pro (Preview) | 74.1 | #64 |
| 3 | Claude Fable 5 | 61.2 | #20 |
| 4 | GPT-5.6 Sol | 59.8 | #22 |
| 5 | Claude Opus 4.8 | 43.8 | #39 |
| 6 | Qwen 3.7 Plus | 36.6 | #98 |
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-vlmsarebiased · How It Works · Data refreshed daily, snapshot 2026-11-05.