PM-LLM-Benchmark: leaderboard
Process mining prompts that supply event logs and process models as text, covering contextual understanding, conformance checking, model generation, querying, hypothesis generation, fairness and optimization. A judge model grades each of the 51 answers from 1 to 10 and the score is their sum divided by ten, so it tops out near 51 and higher is better. A diagram-reading category exists upstream but is excluded here.
Metric: Score. Source: github.com. Saturation forecast: Around 2033. 171 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (xHigh) | 40.2 | #19 (GPT-5.6 Sol) |
| 2 | GPT-5.6 Sol (High) | 39.8 | #19 (GPT-5.6 Sol) |
| 3 | Kimi K3 | 39.2 | #30 |
| 4 | GPT-5.6 Terra (High) | 38.4 | #41 (GPT-5.6 Terra) |
| 5 | Kimi K2.6 | 37.8 | #104 |
| 6 | GPT-5.4 (xHigh) | 37.8 | #79 (GPT-5.4) |
| 7 | GPT-5.5 (High) | 37.7 | #31 (GPT-5.5) |
| 8 | DeepSeek V4 Flash (0731) | 37.7 | #78 |
| 9 | GPT-5.5 Pro | 37.7 | #17 |
| 10 | GPT-5.3 Codex (xHigh) | 37.3 | #69 (GPT-5.3 Codex) |
| 11 | GPT-5.6 Luna (High) | 37.2 | #68 (GPT-5.6 Luna) |
| 12 | GPT-5.4 (High) | 37 | #79 (GPT-5.4) |
| 13 | GPT-5.5 (xHigh) | 36.9 | #31 (GPT-5.5) |
| 14 | GPT-5.2 (xHigh) | 36.8 | #110 (GPT-5.2) |
| 15 | DeepSeek V4 Pro | 36.4 | #102 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=pm-llm-benchmark · How It Works · Data refreshed daily, snapshot 2026-10-20.