ProOPF-B (Few-Shot) - Level 3: leaderboard
Metric: Objective-value accuracy (%) on the 29 Level 3 cases (structural extensions such as new variables, constraints or objective terms, with explicitly stated parameters); ProOPF-B's 121 expert-annotated optimal power flow modeling cases (natural-language operating requirements turned into executable MATPOWER code); a case is correct when the generated implementation executes and its optimal objective value matches the expert implementation's within a small tolerance; few-shot prompting; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 37.93 | #138 |
| 2 | Gemini 3 Pro | 31.03 | #77 |
| 3 | DeepSeek R1 | 17.24 | #245 |
| 4 | GPT-5.2 | 10.34 | #105 |
| 5 | DeepSeek V3.2 | 6.9 | #198 |
| 6 | GPT-5.1 | 6.9 | #131 |
| 7 | Claude 3.5 Sonnet | 0 | #337 |
| 8 | Qwen 3 30B A3B Instruct | 0 | #460 |
| 9 | Qwen3 Coder | 0 | #285 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=proopf-b-few-shot-level-3 · How It Works · Data refreshed daily, snapshot 2026-10-11.