Change8

v0.20.0-rc1

📦 ollamaView on GitHub →
2 features🐛 1 fixes🔧 1 symbols

Summary

This release introduces the new Gemma 4 model family variants (E2B, E4B, 26B, 31B) and enhances tokenizer capabilities with SentencePiece-style BPE support.

✨ New Features

  • Introduced Gemma 4 models with Effective 2B (E2B), Effective 4B (E4B), 26B (MoE with 4B active parameters), and 31B (Dense) variants available via `ollama run`.
  • Added SentencePiece-style BPE support to the tokenizer.

🐛 Bug Fixes

  • The MLX pipeline now correctly respects the tokenizer's `add_bos_token` setting.

Affected Symbols