Importance 2 of 5Covered by 1 source
Google updates small Gemma 4 tokenizers to fix nonsense output bug
Google updated tokenizers for Gemma 4 E4B, E2B and 12B to fix an issue that caused nonsense output on certain tokens. It is unclear if community GGUF builds include the fix.
Why it matters
Teams running these Gemma 4 models should refresh tokenizer files to avoid faulty outputs.