PM-Bench (Hierarchical Multi-Agent): 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; coordinator with three specialist subagents whose suggested state queries are all executed; one run per model). Source: arxiv.org. Saturation forecast: Around March 2027. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Mistral-Large-3-675B-Instruct-2512 | 58.1 |
| 2 | GPT-5.4 | 56.5 |
| 3 | GPT-5.3 Codex | 55.9 |
| 4 | Llama 3.3 70B Instruct | 50.9 |
| 5 | Qwen 3 14B | 43.6 |
| 6 | Qwen 3 32B | 37.5 |
| 7 | Mistral Small 3.2 | 33.6 |
| 8 | Qwen 3 8B | 25.7 |
Interactive version: theaggregate.ai/benchmark?slug=pm-bench-hierarchical-multi-agent · How It Works · Data refreshed daily, snapshot 2026-09-29.