VASPBench (AutoDFT-OpenLoop): leaderboard
VASPBench (Yang et al., 2026): 34 density functional theory calculation tasks from the official VASP documentation, scored by the percentage whose outputs converge and pass validation. Here the LLM writes the whole workflow in one planning call and a deterministic engine executes, monitors and applies rule-based fixes (AutoDFT-OpenLoop); one run per task, higher is better.
Metric: Task success rate (%; converged outputs passing the validation checks on 34 DFT tasks, one up-front LLM plan then deterministic execution, one run per task). Source: arxiv.org. Saturation forecast: Estimated already saturated. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 82.4 | #172 |
| 2 | Gemini 3.1 Pro (Preview) | 76.5 | #91 |
| 3 | Claude Sonnet 4.6 | 76.5 | #158 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=vaspbench-autodft-openloop · How It Works · Data refreshed daily, snapshot 2026-11-12.