Aura Medium GGUF
Portable local inference for llama.cpp-compatible desktops, laptops, and capable edge systems.
View model ↗Medium tier · Two deployment formats
Aura Medium, with more room to think.
Aura Medium is a Gemma 4 12B IT model family built around a larger local foundation. Choose the GGUF or BF16 release without changing the identity at the center of the experience.
Local deployment
GGUF · BF16
Medium tier
Gemma 4 12B IT
Aura Medium is the locally deployed companion you never expected. The medium model is designed to run on all levels of personal computer hardware.
Choose a format
One Aura Medium, matched to your local runtime
Each release is built from the Gemma 4 12B IT model family and validated independently for its target runtime.
Aura Medium BF16
High-fidelity Transformers inference and evaluation for GPUs with sufficient memory.
View model ↗
Shared foundation. The GGUF and BF16 releases share the Aura Medium identity on top of the Gemma 4 12B IT model family, but their execution formats use different numerical representations. Behavioral validation is performed separately for every release.