YOMI-Bench - Kanji Reading Prediction (Multiple Readings): leaderboard
Generate the reading of a Japanese word written in kanji, for kanji with several possible readings, from YOMI-Bench (2026). Exact-match accuracy in percent, averaged within each kanji group and over five paraphrased prompts; higher is better.
Metric: Accuracy (%; exact match of the generated reading of a word, extracted by regular expression, averaged within each kanji group; 120 items on kanji with several readings; mean over five paraphrased prompts; accuracy as a fraction x100). Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 96.78 | #167 |
| 2 | Claude Sonnet 4.5 | 92.16 | #250 |
| 3 | Gemini 2.5 Flash | 86.31 | #453 |
| 4 | GPT-4o | 69.43 | #741 |
| 5 | Mistral Medium 3.1 | 54.1 | #852 |
| 6 | Llama-3.1-Swallow-8B-Instruct-v0.5 | 51.16 | #2704 |
| 7 | Ministral-8B-Instruct-2410 | 30.23 | #4684 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=yomi-bench-kanji-reading-prediction-multiple-readings · How It Works · Data refreshed daily, snapshot 2026-11-13.