LLM Debate Benchmark
Chat & Writing · 2026-03-22
LLM debate benchmark in which models debate both sides of contentious topics and are rated through pairwise outcomes. The leaderboard reports Bradley-Terry ratings.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 1757 |
| Kimi K3 | 1741 |
| Opus 4.7 | 1685 |
| Muse Spark 1.1 | 1684 |
| GPT-5.6 Sol | 1681 |
| Opus 4.8 | 1668 |
| Sonnet 5 | 1619 |
| Sonnet 4.6 | 1597 |
| GLM-5.2 | 1595 |
| GPT-5.4 | 1584 |
| GPT-5.5 | 1564 |
| GLM-5.1 | 1555 |
| Kimi K2.6 | 1547 |
| MiniMax M3 | 1537 |
| Gemini 3.1 Pro Preview | 1529 |
| MiMo-V2.5-Pro | 1526 |
| Grok 4.5 | 1519 |
| Qwen3.6-Max-Preview | 1511 |
| Kimi K2.5 | 1490 |
| Doubao Seed 2.0 Pro | 1489 |
| DeepSeek-V4-Pro | 1478 |
| Qwen3.7-Max | 1470 |
| MiniMax M2.7 | 1469 |
| Grok 4.20 | 1448 |
| Gemini 3.5 Flash | 1444 |
| MiMo-V2-Pro | 1428 |
| Qwen3.5 397B A17B | 1426 |
| Step 3.7 Flash | 1421 |
| ERNIE 5.1 | 1419 |
| Hy3 preview | 1417 |
| Grok 4.3 | 1406 |
| DeepSeek-V3.2 | 1399 |
| Mistral Medium 3.5 | 1382 |
| Gemini 3.1 Flash-Lite Preview | 1366 |
| gpt-oss-120b | 1307 |
| ERNIE 5.0 | 1283 |
| Mistral Large 3 675B Instruct 2512 | 1255 |
| Llama 4 Maverick | 1071 |