Elimination Game (Lechmazur) — leaderboard
20 Questions-style elimination game where models must identify a target from a large set using yes/no questions. TrueSkill ranking from multi-game tournaments across 60+ models.
Metric: TrueSkill μ. Source: github.com. Status: saturation imminent. 61 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 (Medium) | 7.52 |
| 2 | GPT-5 (Medium) | 5.97 |
| 3 | GPT-5 Mini (Medium) | 5.73 |
| 4 | Gemini 3 Flash (Preview) | 5.66 |
| 5 | GPT-4o (Mar 2025) | 5.5 |
| 6 | DeepSeek R1 0528 | 5.35 |
| 7 | Claude Opus 4.1 (Non-reasoning) | 5.21 |
| 8 | Claude Sonnet 4.5 (Thinking 16K) | 5.19 |
| 9 | Grok 4 | 5.15 |
| 10 | GPT-4.5 (Preview) | 5.12 |
| 11 | Claude 3.5 Sonnet (20241022) | 5.1 |
| 12 | Grok 3 | 5.1 |
| 13 | Gemini 3 Pro (Preview) | 4.89 |
| 14 | Claude 3.7 Sonnet | 4.78 |
| 15 | Gemini 2.5 Flash | 4.73 |
Interactive version: theaggregate.ai/benchmark?slug=elimination-game-lechmazur · How the rankings work · Data refreshed daily, snapshot 2026-07-22.