LLM Debate Benchmark
Models argue assigned sides of a proposition against each other over several turns; a blend of judge models picks the winner and the results are rated with Bradley-Terry.
| Category | General |
|---|---|
| Metric | Elo |
| Source | Lech Mazur (original) |
| License | None stated |
| Benchmark page | github.com |
| Status | Scored |
| Weight | 0.8 within General |
| Fit | Discrimination 1.00, noise 0.22 |
Results
| # | Model | Provider | Result | Date |
|---|---|---|---|---|
| 1 | Claude Fable 5.1 High | Anthropic | 1,727 | 2026-10-26 |
| 2 | Claude Fable 5 High | Anthropic | 1,719 | 2026-10-26 |
| 3 | Claude Opus 5 High | Anthropic | 1,717 | 2026-10-26 |
| 4 | Claude Opus 5.5 High | Anthropic | 1,692 | 2026-10-26 |
| 5 | Kimi K3 | Moonshot AI | 1,691 | 2026-10-26 |
| 6 | Gemini 4 Argon High | 1,669 | 2026-10-26 | |
| 7 | Muse Wonder High | Meta | 1,647 | 2026-10-26 |
| 8 | Claude Opus 4.7 High | Anthropic | 1,641 | 2026-10-26 |
| 9 | GPT-5.6 Sol High | OpenAI | 1,637 | 2026-10-26 |
| 10 | GLM-5.3 High | Zhipu AI | 1,634 | 2026-10-26 |
| 11 | Muse Spark 1.1 High | Meta | 1,631 | 2026-10-26 |
| 12 | Claude Opus 4.8 High | Anthropic | 1,624 | 2026-10-26 |
| 13 | Grok 4.7 High | xAI | 1,614 | 2026-10-26 |
| 14 | GPT-6 Astra High | OpenAI | 1,604 | 2026-10-26 |
| 15 | Kimi K3.1 Pro High | Moonshot AI | 1,603 | 2026-10-26 |
| 16 | MiMo-V2.6-Pro Thinking | Xiaomi | 1,600 | 2026-10-26 |
| 17 | Grok 4.6 High | xAI | 1,598 | 2026-10-26 |
| 18 | DeepSeek V4.1 Flash 1017 High | DeepSeek | 1,593 | 2026-10-26 |
| 19 | Muse Spark 1.3 High | Meta | 1,590 | 2026-10-26 |
| 20 | Claude Sonnet 5 High | Anthropic | 1,580 | 2026-10-26 |
| 21 | DeepSeek V4.1 Flash High | DeepSeek | 1,577 | 2026-10-26 |
| 22 | Hy4 preview | Tencent | 1,577 | 2026-10-26 |
| 23 | DeepSeek V4 Pro 0813 High | DeepSeek | 1,565 | 2026-10-26 |
| 24 | GLM-5.2 Max | Zhipu AI | 1,556 | 2026-10-26 |
| 25 | Claude Sonnet 4.6 High | Anthropic | 1,553 | 2026-10-26 |
| 26 | Qwen3.8 Max | Alibaba | 1,546 | 2026-10-26 |
| 27 | GPT-5.4 High | OpenAI | 1,540 | 2026-10-26 |
| 28 | Claude Sonnet 4.6 Non-reasoning | Anthropic | 1,534 | 2026-10-26 |
| 29 | GPT-5.5 High | OpenAI | 1,520 | 2026-10-26 |
| 30 | GLM-5.1 | Zhipu AI | 1,511 | 2026-10-26 |
| 31 | Gemini 3.8 Flash High | 1,505 | 2026-10-26 | |
| 32 | Kimi K2.6 | Moonshot AI | 1,504 | 2026-10-26 |
| 33 | MiniMax-M3 | MiniMax | 1,496 | 2026-10-26 |
| 34 | GPT-5.4 Non-reasoning | OpenAI | 1,487 | 2026-10-26 |
| 35 | Grok 4.5 High | xAI | 1,483 | 2026-10-26 |
| 36 | Gemini 3.1 Pro Preview | 1,482 | 2026-10-26 | |
| 37 | MiMo-V2.5-Pro | Xiaomi | 1,477 | 2026-10-26 |
| 38 | Qwen3.6 Max Preview | Alibaba | 1,467 | 2026-10-26 |
| 39 | Mistral Large 4 High | Mistral | 1,448 | 2026-10-26 |
| 40 | Gemini 3.7 Flash High | 1,448 | 2026-10-26 | |
| 41 | Seed 2.0 Pro | ByteDance | 1,447 | 2026-10-26 |
| 42 | Kimi K2.5 Thinking | Moonshot AI | 1,445 | 2026-10-26 |
| 43 | Qwen3.7 Max | Alibaba | 1,427 | 2026-10-26 |
| 44 | MiniMax-M2.7 | MiniMax | 1,426 | 2026-10-26 |
| 45 | Grok 4.20 (Reasoning) | xAI | 1,404 | 2026-10-26 |
| 46 | Gemini 3.5 Flash | 1,400 | 2026-10-26 | |
| 47 | Grok 4.20 (Non-Reasoning) | xAI | 1,396 | 2026-10-26 |
| 48 | MiMo-V2-Pro | Xiaomi | 1,382 | 2026-10-26 |
| 49 | Qwen3.5 397B-A17B | Alibaba | 1,382 | 2026-10-26 |
| 50 | Hy3 preview | Tencent | 1,379 | 2026-10-26 |
| 51 | Step 3.7 Flash High | StepFun | 1,374 | 2026-10-26 |
| 52 | Grok 4.3 | xAI | 1,363 | 2026-10-26 |
| 53 | DeepSeek V3.2 | DeepSeek | 1,355 | 2026-10-26 |
| 54 | Mistral Medium 3.5 High | Mistral | 1,333 | 2026-10-26 |
| 55 | Gemini 3.1 Flash Lite Preview | 1,322 | 2026-10-26 | |
| 56 | GPT OSS 120B | OpenAI | 1,263 | 2026-10-26 |
| 57 | Mistral Large 3 | Mistral | 1,181 | 2026-10-26 |
| 58 | Llama 4 Maverick | Meta | 1,021 | 2026-10-26 |