The Wrong Kind of Right — leaderboard

Metric: MAR Disability (self-reported). Source: benchmarklist.com. 26 models tracked.

Top models

#ModelScore
1GPT-5.448.8
2Qwen 2.5 72B Instruct38.6
3Llama 3.1 8B28.2
4Mistral 7B26.8
5Llama 3.1 70B26.5
6Grok 4.2025.3
7Qwen 3.5 9B23.9
8Claude Sonnet 4.623.1
9Gemini 3.1 Flash Lite22.8
10Qwen 2.5 7B22.6
11GPT-5.4 Nano21.5
12Gemma 3 27B18.5
13Llama 3.2 3B18.2
14Qwen 3 14B16.4
15GPT-5.4 Mini15.1

Interactive version: theaggregate.ai/benchmark?slug=the-wrong-kind-of-right · How the rankings work · Data refreshed daily, snapshot 2026-07-22.