List your company

Tokenizer Compression on C4 fr

3.34Bits per Token (B/T)Qwen 3
3.31443.48723.663.8328Feb 19, 2026

Evaluation results

MethodLinks
Qwen 3Vocabulary Size=152k2026.023.34
Llama 3Vocabulary Size=128k2026.023.49
DeepSeek R1Vocabulary Size=128k2026.023.53
GPT-OSSVocabulary Size=200k2026.023.97
TrinityVocabulary Size=200k2026.023.98