MacVisor vs. UTM

macOS (installers for macOS 14, 15, 26, and 27, or your own IPSW) and ARM64 Linux.

Windows, Linux, macOS, and many other systems.

No. Guests run on Apple's Virtualization framework.

Yes. QEMU can emulate x86/x64 and many non-ARM architectures.

A signed catalogue of vendor cloud images (Ubuntu, Debian, Fedora, Rocky Linux, AlmaLinux, CentOS Stream, openSUSE Leap, Oracle Linux, Alpine), plus import of your own qcow2, raw, or xz images and .ipsw installers. cloud-init is a settings page in the app and a set of create flags in the CLI: account, hostname, SSH keys, user-data, container runtime.

A gallery of ready-made VM configurations and installs from ISO; a vendor cloud-image catalogue or cloud-init provisioning is not a documented feature.

containerd, Docker, Podman, or k3s, opt-in in the New VM form or with --runtime (none by default). mvz docker <vm> forwards the VM's Docker or Podman API to a socket on the Mac through the guest agent with no network setup, and creates a docker context for it.

Not a documented feature; containers run inside a guest you set up yourself.

Zero-touch installation: cloud-init installs the agent in Linux VMs, and on macOS 27 guests MacVisor creates the account and installs the agent itself over SSH on the first boot. macOS 26 and earlier guests install it by hand from the attached volume.

Guest tools (SPICE and QEMU agents) are installed inside the guest by the user, by backend.

A background service keeps VMs running after the app quits, and each VM runs in its own isolated runner process. Autostart at login and headless starts for CI.

Native Mac GUI with documented headless operation.

One runner process per VM. Quitting the app, or restarting MacVisorService, does not stop running VMs.

VMs run inside the UTM app process, and UTM prompts before quitting while VMs are running; independence from the app process is not something UTM documents.

The mvz CLI has full parity with the app (create, shell, exec, lifecycle, snapshots, clones, networks, ports, files, tasks, service), with --json output and documented exit codes.

utmctl, AppleScript, and Shortcuts; UTM notes that not every feature is exposed.

Disk-only and live snapshots of running Linux and macOS guests, suspend and resume with saved state, pause, VMs shut down or suspended to disk when the Mac logs out or restarts (your choice), and revert with an optional safety snapshot.

The Apple Virtualization backend has no snapshots; disposable mode and VM duplication are documented.

Templates and instant linked clones on ASIF disk layers (a clone shares a sealed base and stores only its own writes), with flatten and trim.

Full VM copies, configuration-only clones, and disposable runs.

Several storage locations on internal, external, or network volumes, with cold migration of stopped VMs and warm migration that suspends, moves, and resumes a running VM in one step.

One VM library; moving VMs between volumes is a manual file operation.

Default NAT and bridging, plus named shared (NAT) and host-only networks; every VM is reachable by name as <vm>.mvz.

Shared, host-only, and bridged modes are available. Named host networks are QEMU-only and do not provide DHCP.

Creates host VLAN interfaces (tags 1–4094) and bridges VM NICs or shared-network uplinks to them.

UTM's “Emulated VLAN” is a userspace QEMU segment; built-in creation of a host 802.1Q VLAN interface is not documented or independently tested.

Per-network DHCP with MAC-to-IPv4 reservations (mvz networks reserve), within vmnet's limits on macOS 27: a shared network re-picks its subnet when it is recreated, and a host-only network with a fixed subnet gets no DHCP at all.

QEMU emulated networking exposes DHCP ranges, but named host networks have no DHCP and MAC address reservations are not documented.

TCP/UDP NAT rules on shared networks, plus live per-VM TCP and UDP tunnels through the guest agent; cloud-image VMs forward the ports they listen on automatically.

TCP/UDP forwarding is documented for QEMU Emulated VLAN mode, not as a rule set on a named shared network.

Host-enforced IPv4 pf policy per network: inbound and outbound defaults with ordered allow/deny rules by protocol, CIDR, and port, set from the app or mvz networks firewall. IPv4 only; traffic between VMs on the same segment is not filtered.

An equivalent per-network host firewall policy is not documented or independently tested.

Per-network DHCP, DNS proxy, NAT44, NAT66, and IPv6 router-advertisement switches, plus IPv4/IPv6 and MTU settings.

QEMU emulated networking exposes subnet, DHCP, and DNS fields, but not an equivalent per-network NAT44/NAT66/IPv6 RA control set.

macOS and Linux agents over vsock carry mvz shell and exec, the per-VM Docker socket, clipboard, bidirectional file transfer, port tunnels, and live IP information; VirtioFS shared folders.

Clipboard, directory sharing, USB sharing, dynamic resolution, and guest tools by backend.

Apple's accelerated graphics device for macOS guests, sharing the host GPU; Linux guests use Virtio graphics.

Virtio GPU acceleration with the Apple Virtualization backend; QEMU guests rely largely on software rendering.

Per-VM CPU, memory, disk, GPU, and network readings, a host dashboard that compares allocation with capacity, a task pane with history, and a log per VM runner.

Desktop VM controls; fleet metrics and storage-location views are not its focus.

Free 30-day trial key. MacVisor 27 Professional ($49, one individual) and Business ($99, organisations) are one-time licenses for one Mac covering every 27.x update; Enterprise starts at $99 per Mac per month, billed annually for 10 or more Macs, with 24x7 support and an SLA, by quote. Every edition has every feature.

Free and open source; an optional paid Mac App Store build supports development.