Ai2 Publishes Bolmo Byte-Level Research in Nature and Releases Bwen 8B and Blama 8B

Ai2's research on Bolmo, a method that converts subword language models to byte-level models, appears in Nature. The institute releases Bwen 8B and Blama 8B, built from Qwen 3 8B and Llama 3 8B, with Bwen 8B beating Bolmo 7B across its evaluation suite. Stage 1 checkpoints keep the original model frozen to speed experimentation.

• Ai2's Bolmo byte-level language model research is published in Nature, and Ai2 releases new checkpoints extending the byteifying approach beyond the Olmo model family.
• Ai2 applies byteifying to Qwen 3 8B and Llama 3 8B, creating Bwen 8B and Blama 8B, with Bwen 8B outperforming Bolmo 7B across the aggregate evaluation suite.
• Ai2 releases Stage 1 checkpoints that keep the original global model frozen while training Bolmo's byte-level components, giving researchers a faster starting point for experimentation.