MHGraphBench - Entity Clustering: leaderboard
Odd-one-out questions over five entities in MHGraphBench (2026), a knowledge-graph-grounded mental-health benchmark built from a curated PrimeKG subgraph, with letter-only multiple-choice answers. Accuracy in percent; higher is better.
Metric: Accuracy (%) on the 2,000 five-entity odd-one-out entity clustering items, letter-only answers; the OpenAI API models answer at temperature 0 (at most 120 completion tokens) with strict answer-letter parsing, the open models are scored by forced-choice option-letter log-probabilities; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 Instant | 92.35 | #472 |
| 2 | GPT-4o | 91.85 | #751 |
| 3 | GPT-4.1 | 91.85 | #505 |
| 4 | GPT-5 Mini | 91.75 | #364 |
| 5 | GPT-5.2 Instant | 90.1 | #304 |
| 6 | Qwen 2.5 32B Instruct | 54.6 | #1247 |
| 7 | Qwen 2.5 7B Instruct | 36.95 | #2723 |
| 8 | Mistral 7B Instruct (v0.3) | 28.15 | #6750 |
| 9 | Llama3-Med42-8B | 24.55 | #5084 |
| 10 | Llama3-OpenBioLLM-8B | 23.35 | #6364 |
| 11 | DeepSeek-R1-Distill-Qwen-7B | 20.4 | #8055 |
| 12 | DeepSeek R1 Distill Qwen 32B | 18.9 | #1834 |
| 13 | Meditron-7B | 18 | #8733 |
| 14 | Llama 3.1 8B Instruct | 17.5 | #4600 |
| 15 | BioMistral-7B | 14.8 | #8332 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=mhgraphbench-entity-clustering · How It Works · Data refreshed daily, snapshot 2026-11-21.