PaliBench - chrF++: leaderboard

PaliBench (2026) asks models to translate 1,700 passages of the Pali Buddhist canon into English, scored against three independent human translations at once. This board reports chrF++, a character n-gram F-score with word bigrams, on a 0-100 scale; higher is better.

Metric: chrF++ (0-100), character n-gram F-score with word bigrams (sacrebleu), 1,700 Pali canon passages translated into English, each scored against three human reference translations (Bodhi, Sujato, Thanissaro), corpus level, models through OpenRouter; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro (Preview)68.5#119
2Gemini 3 Flash (Preview)65.9#152
3Claude Opus 4.565.6#160
4DeepSeek V3.264.1#419
5Kimi K2.561.2#266
6GPT-5.259.1#185
7GLM-4.757.1#380
8Qwen 3 235B A22B 2507 Instruct55.7#616
9Grok 4.1 Fast54.8#427
10Llama 3.3 70B Instruct48.1#1385

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Example tasks

“Dasayimāni, bhikkhave, saṁyojanāni. Katamāni dasa? Pañcorambhāgiyāni saṁyojanāni, pañcuddhambhāgiyāni saṁyojanāni. Katamāni pañcorambhāgiyāni saṁyojanāni? Sakkāyadiṭṭhi, vicikicchā, sīlabbataparāmāso, kāmacchando, byāpādo—imāni pañcorambhāgiyāni saṁyojanāni.

Source

Interactive version: theaggregate.ai/benchmark?slug=palibench-chrf-plus-plus · How It Works · Data refreshed daily, snapshot 2026-11-21.