myaniu/ollama-transfer

ollama-transfer is a command-line tool for moving AI language models between machines that use Ollama, a local AI model runner. If you have downloaded a large model on one computer and need to move it to another without downloading it again, for example to an offline machine or across a slow network, this tool packages the model into a single compressed archive file (.tar.gz) that you can copy via USB drive, scp, or any file transfer method, then load it on the destination machine. The workflow mirrors how Docker handles container images: a save command packs the model, and a load command unpacks it into the Ollama store on the receiving machine. A list command shows what models are available to export, and an inspect command lets you peek inside an archive without actually importing it, which is handy for verifying contents before restoring. The tool is a single static binary with no dependencies, no Python, no separate runtime required. It runs on macOS, Linux, and Windows across both Intel/AMD and ARM processors. When loading, every file in the archive is verified by comparing its SHA-256 checksum against its filename, so corrupted or tampered models are rejected before any data touches your storage. It also guards against path traversal attacks in archives and handles deduplication so blobs shared between model versions are not copied twice. The source code is written in Go and requires Go 1.25 or later to build. Pre-built binaries for all supported platforms are available on the releases page. It is released under the Apache 2.0 license.