llama.cpp code hints at Gemma 4 Unified model

View on Reddit

A merged PR in llama.cpp implements a new 'Gemma 4 Unified' model type, suggesting Google may be preparing a unified model. The code lacks description but shows early implementation details.