MHGraphBench - Two-hop Verification: leaderboard
Metric: Accuracy (%) on the 1,200 two-hop verification items (about half yes), 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 2031. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o | 62.08 |
| 2 | GPT-4.1 | 57.58 |
| 3 | Qwen 2.5 32B Instruct | 50.5 |
| 4 | Llama 3.1 8B Instruct | 50.42 |
| 5 | GPT-5 Mini | 50.25 |
| 6 | GPT-5.2 Instant | 50.17 |
| 7 | GPT-5.1 Instant | 50.17 |
| 8 | Qwen 2.5 7B Instruct | 50.08 |
| 9 | DeepSeek R1 Distill Qwen 32B | 49.75 |
| 10 | DeepSeek-R1-Distill-Qwen-7B | 49.75 |
| 11 | Mistral 7B Instruct (v0.3) | 49.33 |
Interactive version: theaggregate.ai/benchmark?slug=mhgraphbench-two-hop-verification · How It Works · Data refreshed daily, snapshot 2026-10-07.