PM-Bench (Auto Heartbeat 30m): leaderboard
Metric: Set-F1 (%; F1 of the prospective actions the agent selects against the actions actually due, accumulated over the 80 steps and 81 scored tasks of the released simulated week, lures included; single agent with a heartbeat switched on every 30 virtual minutes; one run per model). Source: arxiv.org. Saturation forecast: Around December 2026. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Mistral-Large-3-675B-Instruct-2512 | 74.7 |
| 2 | GPT-5.4 | 74.1 |
| 3 | GPT-5.3 Codex | 71 |
| 4 | Llama 3.3 70B Instruct | 63.4 |
| 5 | Qwen 3 32B | 57.5 |
| 6 | Mistral Small 3.2 | 51.6 |
| 7 | Qwen 3 8B | 39.8 |
| 8 | Qwen 3 14B | 30.3 |
Interactive version: theaggregate.ai/benchmark?slug=pm-bench-auto-heartbeat-30m · How It Works · Data refreshed daily, snapshot 2026-09-29.