GraphInstruct (Generation) - L0 Format Generation: leaderboard
Metric: Level quality score (0-100; the paper's 0-1 score times 100) on the 100 GraphInstruct instructions of level L0, which requires format generation (node and edge counts, syntax); 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 | 95.8 |
| 2 | Qwen 3.5 397B A17B | 95.2 |
| 3 | Qwen 3.5 122B A10B | 94.7 |
| 4 | GPT-4o | 94.2 |
| 5 | Claude Sonnet 4 (20250514) | 93.9 |
| 6 | Qwen 3.5 35B A3B | 92.4 |
| 7 | GPT-4.1 | 92.3 |
| 8 | Llama 3.3 70B Instruct | 90.5 |
| 9 | Llama 3.1 8B Instruct | 89.6 |
| 10 | GPT-4o Mini | 85.2 |
| 11 | GPT-3.5 Turbo | 85 |
Interactive version: theaggregate.ai/benchmark?slug=graphinstruct-generation-l0-format-generation · How It Works · Data refreshed daily, snapshot 2026-10-07.