Ollama recently released version 0.13.1-rc0, introducing several updates for local large language model development. This release addresses specific issues with CUDA VRAM discovery, which should lead to more accurate resource detection and improved utilization for users running models on NVIDIA GPUs. Additionally, the update includes tool calling support for the cogito-v2.1 model, expanding its capability to interact with external functions and services. The nomic-embed-text model will now also default to using Ollama's internal engine, aiming for improved performance and consistency across the platform.