ScienceBoard-Verified - Average: leaderboard

ScienceBoard-Verified scientific computer-use tasks across KAlgebra, Celestia, GRASS GIS, ChimeraX, Lean 4 and TeXstudio; source success rate, with Average the unweighted mean of six applications. Model-based API harness uses the latest three steps plus the current screenshot/accessibility tree; native Grok Build/Codex maintain persistent per-task sessions, with web search disabled. Resource construction and harness identities remain separate run configurations. Average.

Metric: Success rate (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 46 models tracked.

Top models

#ModelScoreOverall rank
1GPT-5.5 (xhigh) (ScienceCUA Run: harness=model-based; resource=Full%20ScienceCUA; evolution=108%20additional%20tasks%3B%20frozen%20library)65.81
2GPT-5.5 (xhigh) (ScienceCUA Run: harness=model-based; resource=SkillCAT; evolution=108%20additional%20tasks%3B%20frozen%20library)63.74
3GPT-5.5 (xhigh) (ScienceCUA Run: harness=model-based; resource=%E2%88%92Knowledge%20addition; evolution=108%20additional%20tasks%3B%20frozen%20library)63.05
4GPT-5.5 (xhigh) (ScienceCUA Run: harness=model-based; resource=Trace2Skill; evolution=108%20additional%20tasks%3B%20frozen%20library)60.46
5GPT-5.5 (xhigh) (ScienceCUA Run: harness=model-based; resource=%E2%88%92Failure%20history; evolution=108%20additional%20tasks%3B%20frozen%20library)60.42
6GPT-5.5 (xhigh) (ScienceCUA Run: harness=Codex; resource=ScienceCUA)59.82
7GPT-5.5 (xhigh) (ScienceCUA Run: harness=model-based; resource=ScienceCUA)57.16
8GPT-5.5 (xhigh) (ScienceCUA Run: harness=Codex; resource=visualskill)56.08
9Grok-4.6 (xhigh) (ScienceCUA Run: harness=Grok%20Build; resource=ScienceCUA)55.41
10GPT-5.5 (xhigh) (ScienceCUA Run: harness=model-based; resource=RAG)54.73

Interactive version: theaggregate.ai/benchmark?slug=scienceboard-verified-average · How It Works · Data refreshed daily, snapshot 2026-11-05.