GraphInstruct (Generation) - L2 Multi-Constraint Composition: leaderboard
Metric: Level quality score (0-100; the paper's 0-1 score times 100) on the 200 GraphInstruct instructions of level L2, which requires four or more joint constraints; zero-shot prompting; five generations per instruction at temperature 0.7 with the InstructGraph code-style prefix; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 93.4 |
| 2 | Qwen 3.5 122B A10B | 91.1 |
| 3 | Claude Sonnet 4 (20250514) | 90 |
| 4 | Qwen 3.5 397B A17B | 88.8 |
| 5 | Qwen 3.5 35B A3B | 87.2 |
| 6 | Llama 3.3 70B Instruct | 86.5 |
| 7 | GPT-4.1 | 86.5 |
| 8 | GPT-4o | 84.3 |
| 9 | GPT-3.5 Turbo | 60.8 |
| 10 | Llama 3.1 8B Instruct | 59.7 |
| 11 | GPT-4o Mini | 58.5 |
Interactive version: theaggregate.ai/benchmark?slug=graphinstruct-generation-l2-multi-constraint-composition · How It Works · Data refreshed daily, snapshot 2026-10-07.