SDABench - Mechanistic: leaderboard

Metric: Accuracy (%) of open-ended answers on the held-out SDA-Synth test split (the 400 mechanistic analysis instances; zero-shot, JSON reasoning and answer fields, the answer scored by relative-error numeric, exact or structured matching with a GPT-4o fallback judge); higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 13 models tracked.

Top models

#ModelScore
1Kimi K2.570
2Qwen 3.5 397B A17B48
3GPT-5.447
4Claude Sonnet 4.645
5DeepSeek V3.241
6Gemini 3.1 Pro (Preview)35.75
7GLM-530.25
8DeepSeek R123.25
9Qwen 3 235B A22B 2507 Instruct22.5
10GPT-5.4 Mini18.75
11Llama 3.3 70B Instruct2.75
12Llama 3.1 8B Instruct0

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