GraphInstruct (Generation) - L1 Single Constraint: leaderboard

GraphInstruct (Generation) L1 Single Constraint scores zero-shot graph generation on the 200 GraphInstruct (2026) instructions whose graphs must satisfy a single explicit constraint (graph type). The level quality score, 0 to 100 and higher is better, is computed by code from validity, distribution match, reference similarity and constraint satisfaction.

Metric: Level quality score (0-100; the paper's 0-1 score times 100) on the 200 GraphInstruct instructions of level L1, which requires a single explicit constraint (graph type); 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. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude Sonnet 4 (20250514)98.3#405
2GPT-4o98.2#751
3Qwen 3.5 122B A10B98#346
4Qwen 3.5 397B A17B97.9#264
5Claude Sonnet 4.697.7#173
6GraphInstruct DeepSeek-V3 (checkpoint unspecified)97.7
7GPT-4.196.8#505
8Qwen 3.5 35B A3B95.5#492
9Llama 3.3 70B Instruct94.8#1385
10GPT-3.5 Turbo90.5#3005
11GPT-4o Mini89.8#1667
12Llama 3.1 8B Instruct73#4600

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=graphinstruct-generation-l1-single-constraint · How It Works · Data refreshed daily, snapshot 2026-11-21.