pfgen-bench - Completion Mode - Truthfulness: leaderboard

Metric: Truthfulness Score. Source: github.com. 314 models tracked.

Top models

#ModelScore
1DeepSeek V30.97
2Qwen 3 235B A22B (Non-reasoning)0.95
3plamo-2-8B0.94
4sarashina2-70B0.92
5c4ai-command-r-plus0.91
6Nemotron-4 340B Instruct0.91
7sarashina2.2-3B0.91
8Mistral Large 2 (Nov) Instruct (2411)0.9
9sarashina2-13B0.9
10sarashina2.2-3B-instruct-v0.10.89
11Llama 3.1 405B0.89
12llm-jp-3-13B0.88
13Qwen 3 32B (Non-reasoning)0.88
14sarashina2-7B0.87
15Phi-40.86

Interactive version: theaggregate.ai/benchmark?slug=pfgen-bench-completion-mode-truthfulness · How It Works · Data refreshed daily, snapshot 2026-09-19.