pfgen-bench - Chat Mode - Score: leaderboard
Metric: pfgen Score (mean of three). Source: github.com. 35 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4 (20250514) (Thinking) | 0.95 |
| 2 | MiniMax-Text-01 | 0.83 |
| 3 | Qwen 2.5 72B Instruct | 0.78 |
| 4 | Mistral Large 2 (Nov) Instruct (2411) | 0.73 |
| 5 | QwQ 32B-Preview | 0.64 |
| 6 | Qwen 2.5 32B Instruct | 0.63 |
| 7 | Phi-4 | 0.63 |
| 8 | sarashina2.2-3B-instruct-v0.1 | 0.62 |
| 9 | Qwen 2.5 14B Instruct | 0.61 |
| 10 | Qwen 3 8B (Non-reasoning) | 0.59 |
| 11 | Mistral Small 3 | 0.55 |
| 12 | Qwen 2.5 7B Instruct | 0.5 |
| 13 | TinySwallow-1.5B-Instruct | 0.49 |
Interactive version: theaggregate.ai/benchmark?slug=pfgen-bench-chat-mode-score · How It Works · Data refreshed daily, snapshot 2026-09-19.