From the source
IBM's Granite Switch 4.1 models (8B and 30B parameters) integrate a Decision aLoRA for general-purpose decision-making, a Guardian aLoRA for safety-related decisions, and the Granite-Speech-5.0-Turbo-CTC ASR model into a single hosted endpoint.
The models reuse the base model's KV cache for adapter-based decisions, enabling fast switching between chat, decision, and speech tasks with minimal additional VRAM.
The models are released with predefined adapters, and users can compose custom Granite Switch models using adapters from Granite Libraries.
