SDABench - Inferential (MCQ): leaderboard

Metric: Accuracy (%) of four-option multiple-choice answers with perturbation-based distractors on the held-out SDA-Synth test split (the 400 inferential analysis instances; zero-shot; 25% chance); higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 13 models tracked.

Top models

#ModelScore
1GPT-5.478.5
2GLM-574.5
3Claude Sonnet 4.674.25
4Qwen 3.5 397B A17B73.75
5Kimi K2.573
6Gemini 3.1 Pro (Preview)70.75
7DeepSeek V3.269
8DeepSeek R167
9GPT-5.4 Mini66.75
10Qwen 3 235B A22B 2507 Instruct59.25
11Llama 3.3 70B Instruct39.5
12Llama 3.1 8B Instruct17.75

Interactive version: theaggregate.ai/benchmark?slug=sdabench-inferential-mcq · How It Works · Data refreshed daily, snapshot 2026-09-29.