Diplomacy: Betrayal Tendency: leaderboard

60 critical-state games resumed Fall 1914; betrayal behavior rather than general capability. Retain existing Elo exclusion.

Metric: Betrayal rate (%), 60 critical-state games resumed at Fall 1914. Source: goodstartlabs.com. 17 models tracked.

Top models

#ModelScoreOverall rank
1GPT-50#98
2GPT-OSS-120B0#341
3GPT-5 Mini0#186
4AFM-4.5B0#1372
5Grok 4 Fast0#254
6Gemini 3 Pro11.7#80
7Llama 4 Maverick13.3#464
8Qwen 3 Max21.7#208
9Claude Sonnet 4.531.7#142
10DeepSeek R140#251
11O341.7#124
12Gemini 2.5 Flash51.7#245
13Claude Opus 4.566.7#83
14Kimi K276.7#243
15GLM-4.685#252

Interactive version: theaggregate.ai/benchmark?slug=diplomacy-betrayal-tendency · How It Works · Data refreshed daily, snapshot 2026-10-20.