PGG-Bench (Lechmazur): leaderboard
Metric: TrueSkill μ. Source: github.com. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 03-25) | 14.35 |
| 2 | O1 | 13.4 |
| 3 | Mistral Large 2 (Jul) | 11.93 |
| 4 | Gemini 2.0 Pro (Preview 02-05) | 11.14 |
| 5 | O3 Mini (Medium) | 11.11 |
| 6 | DeepSeek V3 | 10.94 |
| 7 | GPT-4.5 (Preview) | 10.77 |
| 8 | Llama 3.3 70B | 10.19 |
| 9 | Grok 2 (1212) | 10.12 |
| 10 | DeepSeek R1 | 9.66 |
| 11 | Qwen 2.5 Max | 9.46 |
| 12 | Claude 3.7 Sonnet | 9.33 |
| 13 | Llama 3.1 405B | 8.49 |
| 14 | Gemini 2.0 Flash (01-21) (Thinking) | 8.26 |
| 15 | Claude 3.5 Sonnet (20241022) | 8.05 |
Interactive version: theaggregate.ai/benchmark?slug=pgg-bench-lechmazur · How It Works · Data refreshed daily, snapshot 2026-09-19.