pfgen-bench - QA Mode - Fluency: leaderboard
pfgen-bench (Preferred Networks) asks 50 Japanese questions expecting formal answers of about 100 characters and scores them without an LLM judge, averaged over trials; each metric is discounted linearly past 100 characters, to zero at 150. QA mode: an instruction precedes the examples in a chat template's user turn. Fluency is 10-gram overlap with several reference answer sets, scaled so a typical reference scores 1; higher is better.
Metric: Fluency Score. Source: github.com. Saturation forecast: Estimated already saturated. 166 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4 (20250514) (Thinking) | 1 | #164 (Claude Opus 4 (20250514)) |
| 2 | Claude 3.5 Sonnet (20240620) | 0.95 | #308 |
| 3 | GPT-4o | 0.92 | #357 |
| 4 | Claude 3 Opus (20240229) | 0.91 | #428 |
| 5 | GPT-4 | 0.89 | #459 |
| 6 | Qwen 2.5 72B Instruct | 0.87 | #469 |
| 7 | Qwen Max | 0.86 | #388 |
| 8 | MiniMax-Text-01 | 0.85 | #519 |
| 9 | DeepSeek R1 | 0.84 | #262 |
| 10 | Gemini 1.5 Pro (002) | 0.83 | #368 |
| 11 | Mistral Large 2 (Nov) Instruct (2411) | 0.82 | #480 |
| 12 | GPT-4 Turbo | 0.82 | #417 |
| 13 | sarashina2.2-3B-instruct-v0.1 | 0.81 | #1115 |
| 14 | Claude 3 Sonnet (20240229) | 0.81 | #694 |
| 15 | Nemotron-4 340B Instruct | 0.77 | #644 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=pfgen-bench-qa-mode-fluency · How It Works · Data refreshed daily, snapshot 2026-11-05.