HeaRTS: leaderboard
Metric: Overall score (x100): macro-average of the Perception, Inference, Generation and Deduction category scores over HeaRTS's 110 health time-series tasks (16 datasets, 20,226 test cases); each task is scored 0 to 1 by accuracy, IoU or 1 - sMAPE/2 on min-max normalized signals; the model reasons over the signal files by writing and running Python code in a CodeAct agent loop with a minimal package set; higher is better. Source: arxiv.org. Saturation forecast: Around September 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 69 | #54 |
| 2 | GLM-5 (Thinking) | 68 | #137 (GLM-5) |
| 3 | GLM-4.7 (Thinking) | 66 | #185 (GLM-4.7) |
| 4 | Kimi K2 (Thinking) | 65 | #236 (Kimi K2) |
| 5 | Grok 4.1 Fast (Reasoning) | 65 | #208 (Grok 4.1 Fast) |
| 6 | DeepSeek V3.1 | 64 | #260 |
| 7 | GPT-4.1 Mini | 63 | #346 |
| 8 | Claude Haiku 4.5 | 63 | #271 |
| 9 | Qwen 3 Coder 480B A35B Instruct | 63 | #302 |
| 10 | Gemini 2.5 Pro | 62 | #145 |
| 11 | Gemini 2.5 Flash | 62 | #237 |
| 12 | GPT-5 Mini | 62 | #176 |
| 13 | MiniMax-M2 | 61 | #307 |
| 14 | Llama 4 Maverick | 58 | #451 |
| 15 | Nemotron Nano 12B V2 | 47 | #656 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=hearts · How It Works · Data refreshed daily, snapshot 2026-10-11.