Allen Institute for AI Unveils Bolmo Models That Process Raw Text Bytes Without Tokenizers

Allen Institute for AI launches Bolmo, revolutionary language models that process raw UTF-8 bytes directly without tokenizers, offering superior multilingual capabilities and handling of text errors through innovative 'bytefying' technology that transforms existing models cost-effectively.

• Allen Institute for AI introduces Bolmo, a new family of byte-level language models including 7B and 1B versions that operate directly on raw UTF-8 bytes without requiring tokenizers
• Bolmo models are built by 'bytefiying' existing Olmo 3 models through a two-stage training process, making byte-level training more cost-effective than training from scratch
• The models demonstrate strong performance against comparable byte-level and character-based models while offering enterprises better handling of multilingual text, misspellings, and noisy inputs