MHGraphBench - Relation Prediction: leaderboard
Four-way drug-disease relation questions (indication, contraindication, off-label use or none) 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 1,634 four-way drug-disease relation prediction items (indication, contraindication, off-label use, none), 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: Around February 2027. 15 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 Instant | 58.63 | #304 |
| 2 | GPT-5.1 Instant | 58.08 | #472 |
| 3 | GPT-5 Mini | 57.28 | #364 |
| 4 | GPT-4.1 | 54.96 | #505 |
| 5 | GPT-4o | 53.55 | #751 |
| 6 | Qwen 2.5 32B Instruct | 38.43 | #1247 |
| 7 | DeepSeek-R1-Distill-Qwen-7B | 32.62 | #8055 |
| 8 | Llama3-Med42-8B | 27.97 | #5084 |
| 9 | Qwen 2.5 7B Instruct | 25.89 | #2723 |
| 10 | Mistral 7B Instruct (v0.3) | 25.7 | #6750 |
| 11 | Meditron-7B | 25.34 | #8733 |
| 12 | Llama3-OpenBioLLM-8B | 23.68 | #6364 |
| 13 | Llama 3.1 8B Instruct | 23.5 | #4600 |
| 14 | DeepSeek R1 Distill Qwen 32B | 22.46 | #1834 |
| 15 | BioMistral-7B | 20.99 | #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-relation-prediction · How It Works · Data refreshed daily, snapshot 2026-11-21.