FrontierOR Hard - Quality-Time Efficiency: leaderboard
Quality-time efficiency in percent on the 50 most computationally demanding tasks: the share of instances on which programs that LLMs generate in one shot for large-scale optimization problems from operations research papers, run on hidden large instances against an expert-verified Gurobi reference (FrontierOR, 2026) match Gurobi on both solution quality and runtime. Higher is better.
Metric: Quality-time efficiency (QTE) times 100 (%) on the 50-task Hard subset (the most computationally demanding tasks), one-shot generation of a complete optimization program from the natural-language problem and instance schemas (a few self-debug iterations allowed, tiny-instance sanity gate), run on the hidden large instances against an expert-verified Gurobi reference; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 | 32 | #113 |
| 2 | Gemini 3.1 Pro (Preview) | 22 | #99 |
| 3 | GPT-5.3 Codex | 18 | #124 |
| 4 | DeepSeek R1 | 11 | #482 |
| 5 | Qwen 3 Coder Plus | 7 | #720 |
| 6 | Grok 4.20 | 6 | #301 |
| 7 | Llama 4 Maverick | 2 | #1110 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=frontieror-hard-quality-time-efficiency · How It Works · Data refreshed daily, snapshot 2026-11-21.