{
 "measured_at": "2026-10-27",
 "source": "NTREX-128 (Microsoft, CC BY-SA 4.0): newstest2019-src.eng.txt and newstest2019-ref.tha.txt, 1,997 English news sentences and their professional Thai translations, https://github.com/MicrosoftTranslator/NTREX",
 "method": "Each English sentence and its professional Thai translation counted with each tokenizer, no special tokens. corpus_th_over_en = all Thai tokens / all English tokens; th_over_en_median, p10, p90 over per-sentence ratios; tokens per character = total tokens / total characters per language.",
 "sentences": 1997,
 "tokenizers": {
  "gpt_o200k": {
   "th_tok_per_char": 0.4,
   "en_tok_per_char": 0.2093,
   "corpus_th_over_en": 1.8318,
   "th_over_en_median": 1.8333,
   "p10": 1.3541,
   "p90": 2.4069,
   "n": 1997,
   "th_tokens": 94997,
   "en_tokens": 51860
  },
  "typhoon25_qwen3": {
   "th_tok_per_char": 0.5349,
   "en_tok_per_char": 0.2142,
   "corpus_th_over_en": 2.3942,
   "th_over_en_median": 2.3929,
   "p10": 1.7401,
   "p90": 3.1528,
   "n": 1997,
   "th_tokens": 127017,
   "en_tokens": 53052
  },
  "qwen36": {
   "th_tok_per_char": 0.2612,
   "en_tok_per_char": 0.2133,
   "corpus_th_over_en": 1.1743,
   "th_over_en_median": 1.1591,
   "p10": 0.8947,
   "p90": 1.5217,
   "n": 1997,
   "th_tokens": 62038,
   "en_tokens": 52831
  }
 }
}