Multi-turn Debate (Lechmazur): leaderboard
Adversarial multi-turn debate benchmark where LLMs argue opposing sides of policy propositions. Each matchup runs twice with sides reversed. Ranked via Bradley-Terry ratings across 1,000+ debates judged by LLMs. Tests factual accuracy under pressure, counterargument quality, and epistemic discipline.
Metric: Bradley-Terry Rating. Source: github.com. Saturation forecast: Around November 2027. 59 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5.1 (High) | 1733.5 | #9 (Claude Fable 5.1) |
| 2 | Claude Fable 5 (High) | 1722.4 | #16 (Claude Fable 5) |
| 3 | Claude Opus 5 (High) | 1720.1 | #12 (Claude Opus 5) |
| 4 | Kimi K3 | 1694 | #30 |
| 5 | Claude Opus 5.5 (High) | 1691.1 | #3 (Claude Opus 5.5) |
| 6 | Gemini 4 Argon (High) | 1672.7 | #8 (Gemini 4 Argon) |
| 7 | Claude Opus 4.7 (High) | 1644.1 | #49 (Claude Opus 4.7) |
| 8 | GPT-5.6 Sol (High) | 1642 | #19 (GPT-5.6 Sol) |
| 9 | GLM-5.3 (High) | 1637.7 | #39 (GLM-5.3) |
| 10 | Muse Spark 1.1 (High) | 1633.9 | #52 (Muse Spark 1.1) |
| 11 | Claude Opus 4.8 (High) | 1626.7 | #37 (Claude Opus 4.8) |
| 12 | Grok 4.7 (High) | 1616.1 | #25 (Grok 4.7) |
| 13 | GPT-6 Astra (High) | 1608 | #6 (GPT-6 Astra) |
| 14 | Grok 4.6 (High) | 1600.9 | #26 (Grok 4.6) |
| 15 | Muse Spark 1.3 (High) | 1593.3 | #24 (Muse Spark 1.3) |
No result here: #7 GPT-6.1 Sol, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=multi-turn-debate-lechmazur · How It Works · Data refreshed daily, snapshot 2026-10-20.