Buyout Game (Lechmazur): leaderboard
Multi-agent bargaining/elimination game where 8 LLMs play with starting balances, transfers, votes, and a buyout endgame. Ranked by Bradley-Terry ratings over wealth-based pairwise results from mirrored match packs.
Metric: Bradley-Terry Rating. Source: github.com. Saturation forecast: Around July 2027. 36 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (High) | 1975.4 | #31 (GPT-5.5) |
| 2 | GPT-5.4 (High) | 1960.6 | #79 (GPT-5.4) |
| 3 | GLM-5 | 1791.6 | #141 |
| 4 | Claude Opus 4.6 (Thinking, High) | 1759.7 | #63 (Claude Opus 4.6) |
| 5 | DeepSeek V4 Pro | 1681.6 | #102 |
| 6 | Gemini 3.5 Flash | 1667.2 | #60 |
| 7 | GPT-5.4 (Non-reasoning) | 1654 | #79 (GPT-5.4) |
| 8 | Hy3-preview (High) | 1620.9 | #157 (Hy3-preview) |
| 9 | Gemini 3.1 Flash Lite (Preview) | 1615.3 | #194 |
| 10 | GLM-5.1 | 1604.9 | #97 |
| 11 | Gemini 3.1 Pro (Preview) | 1563.8 | #58 |
| 12 | DeepSeek V4 Flash | 1555.8 | #137 |
| 13 | Claude Sonnet 4.6 (Non-reasoning) | 1549.9 | #90 (Claude Sonnet 4.6) |
| 14 | Grok 4.20 0309 (Reasoning) | 1549.4 | #112 (Grok 4.20 0309) |
| 15 | Grok 4.3 | 1525.1 | #188 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=buyout-game-lechmazur · How It Works · Data refreshed daily, snapshot 2026-10-20.