CanMT - Japanese to Chinese: leaderboard

Metric: Translation score (1-7) on the 106 Japanese-to-Chinese CanMT sentences containing culture-specific items, GPT-5-nano judge with the reference translation scoring contextual accuracy, cultural adaptation, functional equivalence, fidelity and naturalness on 7-point Likert scales, averaged per sentence; default translation prompt; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 22 models tracked.

Top models

#ModelScore
1GPT-45.23
2Grok 4.15.17
3DeepSeek V3.25.07
4GPT-4o4.93
5Qwen 2.5 72B Instruct4.93
6DeepSeek R14.84
7Qwen 2.5 32B Instruct4.82
8Qwen 3 32B (Non-reasoning)4.77
9Qwen 3 14B (Non-reasoning)4.73
10Gemini 2.5 Flash Lite4.72
11Qwen 2.5 14B Instruct4.66
12Qwen 3 8B (Non-reasoning)4.65
13Llama 3.3 70B Instruct4.43
14Qwen 2.5 7B Instruct4.28
15Qwen 3 4B (Non-reasoning)4.24

Interactive version: theaggregate.ai/benchmark?slug=canmt-japanese-to-chinese · How It Works · Data refreshed daily, snapshot 2026-10-07.