GABench - Graph Retrieval - Graph Level (Text-Attributed Graphs): leaderboard
Metric: Task success rate (%; share of tasks whose final answer a DeepSeek-V4-Flash judge marks consistent with the verified ground truth; graph-level graph retrieval (node attributes, neighbourhoods, edges, subgraphs and graph statistics) with its 10 retrieval tools on text-attributed graphs (Reddit, Instagram, History, Product, Arxiv and Cora, 200 generated tasks per dataset and category); LLM agent in the OpenClaw harness, one run). Source: arxiv.org. Saturation forecast: Around December 2026. 6 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Pro | 71.97 |
| 2 | GPT-5 Mini | 66.05 |
| 3 | GLM-4.5 Flash | 30.63 |
| 4 | GLM-5-Turbo | 22.73 |
| 5 | Gemini 3 Flash | 15.63 |
| 6 | Qwen 3 235B A22B | 5.48 |
Interactive version: theaggregate.ai/benchmark?slug=gabench-graph-retrieval-graph-level-text-attributed-graphs · How It Works · Data refreshed daily, snapshot 2026-09-29.