Google releases Gemma 4 12B, an open encoder-free multimodal model

Google shipped Gemma 4 12B, a compact multimodal model that handles text and images without separate encoder components. For indie developers, this means easier local deployment and lower resource requirements compared to traditional vision-language architectures—useful if you're building image-aware features without cloud API costs.