A new tokenizer-adaptation method cuts Thai tokens 4× on pretrained models
kastnerkyle · x · 2026-07-23
Researchers describe a practical way to adapt the tokenizer of a pretrained model without retraining from scratch. Instead of adding unused tokens through the usual extension route, they continue BPE learning on new data and pair it with leaf-based vocabulary pruning.
Reported gains include:
- Thai: 4.0× fewer tokens
- Vietnamese: 2.6× fewer tokens
- Hindi: 2.4× fewer tokens
- Estimated 2.2× to 3.7× faster per-character decoding on-device for these languages
The paper argues that tokenizers are under-studied despite strongly affecting multilingual model usage, and the authors release the method as an open-source toolkit.
More from Models
- User Praises DeepSeek's Model as Surprisingly Fast and Good in Hands-on Test — MaziyarPanahi · 2026-09-11
- Qwen3-8B gets a KV-approximation add-on that halves prefill time without touching the model — teortaxesTex · 2026-09-11
- Pro 20x tier burns 60% of weekly quota in under a day with GPT-6 Astra — rschu · 2026-09-11
- Google isn't honoring its own Gemini Grounded Search pricing: only 289 of 15,000+ requests counted as free — ItalyExpat · 2026-09-11
- Is DeepSeek's rumored K3 a scaled-down model, or something bigger? X users debate — teortaxesTex · 2026-09-11
- Unreleased 'GPT 6 Sol' model spotted in OpenAI's API — SteveEricJordan · 2026-09-11