ECG-Reasoning-Benchmark (MIMIC-IV-ECG) - Depth: leaderboard
Metric: Reasoning depth (0 to 4): average number of the four verification steps passed per required finding, ground-truth history supplied after each failure, pooled over findings, on the 3,355 MIMIC-IV-ECG samples, multi-turn ECG reasoning chains for 17 core diagnoses (balanced positive and negative cases), 12-lead ECGs given as plotted images (OpenTSLM as a 100 Hz time series), temperature 0; each step's answer is checked for semantic agreement with the ground truth by a Gemini-3-Flash judge; higher is better. Source: arxiv.org. Saturation forecast: Around February 2028. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 2.07 | #78 |
| 2 | Gemini 2.5 Pro | 1.98 | #145 |
| 3 | GPT-5.2 | 1.93 | #105 |
| 4 | Gemini 2.5 Flash | 1.78 | #237 |
| 5 | GPT-5 Mini (2025-08-07) | 1.7 | #165 |
| 6 | Qwen 3 VL 32B Instruct | 1.59 | #276 |
| 7 | Qwen 3 VL 8B Instruct | 1.54 | #401 |
| 8 | Llama 3.2 90B Vision Instruct | 1.27 | #611 |
| 9 | Llama 3.2 11B Instruct | 1.02 | #1112 |
| 10 | MedGemma-4B-IT | 0.8 | #842 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=ecg-reasoning-benchmark-mimic-iv-ecg-depth · How It Works · Data refreshed daily, snapshot 2026-10-11.