Colima vs SCONE

Colima runs container environments on macOS and Linux, with Docker, containerd, Kubernetes and Incus runtime options. Docker compatibility is intended to let existing Docker workflows and tools continue without changes; containerd users can use nerdctl for a Docker-like command line. It can launch a local, single-node k3s cluster, and users can select an exact K3s version. CPU, memory and disk allocations are configurable, and multiple isolated instances can run with different settings. Colima also documents GPU-accelerated AI workloads on Apple Silicon using Krunkit, with model runners supporting Docker AI Registry and HuggingFace; Ramalama also supports Ollama registries. Those workloads require Apple Silicon, macOS 13 or newer, and Krunkit, with at least 8 GB of RAM recommended. The AI containers are isolated from the host filesystem, other containers and network access unless network access is enabled. Colima is free and MIT-licensed, with installation options for macOS and Linux. Incus virtual machines require m3 or newer Apple Silicon.

SCONE has no write-up on Specifiction yet; its figures are above.