Codesota · Benchmark · Open Graph BenchmarkHome/Leaderboards/Structured Data & Forecasting/Node Classification/Open Graph Benchmark
Stanford University

Open Graph Benchmark.

Large-scale graph ML benchmarks from Stanford. ogbn-arxiv: node classification on 169K CS arXiv papers using citation graph. ogbn-products: node classification on 2.4M Amazon product nodes using co-purchasing graph.

Paper ↗Leaderboard ↓
§ 01 · Leaderboard

Results by metric.

Found a wrong score or missing run?
Use row edits to send a sourced correction into moderation.
Add / edit result ↗Report issue ↗

ogbn-products Accuracy

Test accuracy on ogbn-products (Amazon product node classification, 47 categories)

Higher is better

Trust tiers for ogbn-products Accuracyverifiedpapervendorcommunityunverified

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

RankModelTrustScoreYearLinksFix
01BiGTex
Rank #1 on ogbn-products. BiGTex with external data. Oct 2025.
paper90.292026Source ↗Looks wrong?
02GLEM+GIANT+SAGN+SCR
Rank #2. GLEM+GIANT+SAGN+SCR: graph-language model ensemble with GIANT text features. Oct 2022.
paper87.372026Source ↗Looks wrong?
03LD+GIANT+SAGN+SCR
Rank #3. LD+GIANT+SAGN+SCR: Label Denoising with GIANT+SAGN+SCR. Sep 2023.
paper87.182026Source ↗Looks wrong?
04GraDBERT & RevGAT+KD
Rank #4. GraDBERT+GIANT & SAGN+SLE+CnS: Apr 2023.
paper86.922026Source ↗Looks wrong?
05GraphSAGE
83.89 ± 0.36. OGB paper baseline. Best among vanilla GNNs on ogbn-products.
verified83.892026Source ↗Looks wrong?
06GCN
82.33 ± 0.19. OGB paper baseline.
verified82.332026Source ↗Looks wrong?
07GAT
80.99 ± 0.16. OGB paper baseline.
verified80.992026Source ↗Looks wrong?

ogbn-arxiv Accuracy

Test accuracy on ogbn-arxiv (arXiv CS paper classification, 40 subject areas)

Higher is better

Trust tiers for ogbn-arxiv Accuracyverifiedpapervendorcommunityunverified

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

RankModelTrustScoreYearLinksFix
01BiGTex
Rank #1 on ogbn-arxiv. BiGTex: integrates GNNs and LLMs through stacked Graph-Text Fusion Units. Uses external data. Apr 2025.
paper88.512026Source ↗Looks wrong?
02SimTeG+TAPE+RevGAT
Rank #2. SimTeG+TAPE+RevGAT: fine-tunes LLM as text encoder for graph nodes, combined with RevGAT. Aug 2023.
paper78.032026Source ↗Looks wrong?
03TAPE+RevGAT
Rank #3. TAPE+RevGAT: harnessing LLM explanations as features to boost GNN performance. Published at ICLR 2024.
paper77.52026Source ↗Looks wrong?
04SimTeG+TAPE+GraphSAGE
Rank #4. SimTeG+TAPE+GraphSAGE: SimTeG text encoder with GraphSAGE. Aug 2023.
paper77.482026Source ↗Looks wrong?
05LD+REVGAT
Rank #5. LD+REVGAT: Label Denoising with RevGAT. Sep 2023.
paper77.262026Source ↗Looks wrong?
06GraDBERT & RevGAT+KD
Rank #6. GraDBERT & RevGAT+KD: Graph-BERT with RevGAT knowledge distillation. Apr 2023.
paper77.212026Source ↗Looks wrong?
07GLEM+RevGAT
Rank #7. GLEM+RevGAT: learning on text-attributed graphs via variational inference. ICLR 2023.
paper76.942026Source ↗Looks wrong?
08GCN
73.60 ± 0.18. OGB paper baseline. Verified by NeurIPS 2024 reassessment (arxiv:2406.08993).
verified73.62026Source ↗Looks wrong?
09GAT
73.30 ± 0.16. OGB paper baseline. GAT+C&S reaches 73.86 but this is vanilla GAT.
verified73.32026Source ↗Looks wrong?
10GraphSAGE
72.95 ± 0.31. OGB paper baseline. Verified by NeurIPS 2024 reassessment (arxiv:2406.08993).
verified72.952026Source ↗Looks wrong?
§ 04 · Submit a result

Add to the leaderboard.

← Back to Node Classification