LLM Emergent Collusion — leaderboard
Measures the rate at which frontier LLMs spontaneously form price-fixing cartels in a sealed-bid auction simulation — without being instructed to collude. Tests emergent strategic coordination and anti-competitive behavior.
Metric: Collusion Rate (%). Source: github.com. Status: saturated. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Grok 4 | 75 |
| 2 | O4 Mini | 62 |
| 3 | Qwen 3 235B A22B | 57 |
| 4 | O3 | 51 |
| 5 | Mistral Medium 3 | 47 |
| 6 | Gemini 2.5 Pro | 39 |
| 7 | Claude Opus 4 (Non-reasoning) | 36 |
| 8 | Claude Sonnet 4 (Non-reasoning) | 32 |
| 9 | DeepSeek V3 (0324) | 26 |
| 10 | GPT-4o (Mar 2025) | 23 |
| 11 | Grok 3 | 19 |
Interactive version: theaggregate.ai/benchmark?slug=llm-emergent-collusion · How the rankings work · Data refreshed daily, snapshot 2026-07-22.