FLUX.2-small-decoder and gemma-4-12B-it models now available on Amazon SageMaker JumpStart

This article announces the availability of FLUX.2-small-decoder and gemma-4-12B-it models on Amazon SageMaker JumpStart, expanding the foundation models portfolio for AWS customers.

• FLUX.2-small-decoder delivers 1.4× faster image decoding with 1.4× lower VRAM consumption for FLUX.2 pipelines
• Distilled VAE decoder serves as drop-in replacement with minimal quality loss, ideal for production-grade image generation
• gemma-4-12B-it provides unified multimodal understanding across text, image, and audio inputs
• Features encoder-free architecture with native function calling and agentic workflow support
• Runs on 16GB RAM with performance near Google's larger 26B MoE model at half the memory footprint
• Deploy models via SageMaker console or Python SDK with just a few clicks

These models enable customers to deploy high-performance, scalable AI solutions on AWS infrastructure for specialized use cases.