NPHardEval — leaderboard
Evaluates LLMs on 9 NP-hard computational problems (TSP, graph coloring, knapsack, etc.) spanning P, NP-complete, and NP-hard complexity classes.
Metric: Average Accuracy (%). Source: huggingface.co. Status: saturation imminent. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4 Turbo | 37.74 |
| 2 | Claude 2 | 26.42 |
| 3 | GPT-3.5 Turbo | 25.86 |
| 4 | Claude Instant | 23.94 |
| 5 | Qwen-14B-Chat | 22.28 |
| 6 | Yi 34B (Chat) | 18.93 |
| 7 | Mistral 7B Instruct (v0.1) | 18.22 |
| 8 | Phi-2 | 9.21 |
| 9 | vicuna-13B-v1.3 | 7.56 |
| 10 | Phi-1.5 | 0.22 |
| 11 | mpt-30B Instruct | 0.02 |
Interactive version: theaggregate.ai/benchmark?slug=nphardeval · How the rankings work · Data refreshed daily, snapshot 2026-07-22.