OWE-Bench - Scan to CAD: leaderboard

OWE-Bench - Scan to CAD — Expert-built long-horizon science, software, ML and engineering tasks with task-specific hardware and4–18hour budgets; native baseline/ceiling normalized held-out reward,mean of8attempts and equal task means. Web frozen October1,2026;deterministic sealed verifier. Native model/effort and every attempt retained. Score fraction0–1;higher is better.

Metric: Mean normalized held-out task score (fraction, 0–1). Source: openworldengineer.mercor.com. Saturation forecast: Around October 2027. 14 models tracked. Updated 15 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-6 AstraMax0.4971#6
2GPT-6.1 SolMax0.4965#10
3Claude Fable 5.2Max0.4891#1
4Claude Opus 5.5Max0.3717#5
5Claude Sonnet 5.5Max0.3651#19
6Claude Fable 5.1Max0.3496#15
7Gemini 4 CarbonHigh0.3032#2
8Gemini 4 ArgonHigh0.2873#12
9DeepSeek V4.1 FlashMax0.1389#62
10Muse WonderMax0.1387#28
11GLM-5.3Max0.07683#68
12Grok 4.7xHigh0.07053#47
13Beam 1.0Max0.0677#206
14InklingMax0.002025#281

Interactive version: theaggregate.ai/benchmark?slug=owe-bench-scan-to-cad · How It Works · Data refreshed daily, snapshot 2026-11-16.