Off Grid AI Desktop 如何在 Mac 上更新 llama.cpp 等引擎而毋須重新安裝（2026）

A new model lands on Hugging Face, you download the GGUF, and your Mac will not load it because the llama.cpp build inside your app predates the model's architecture. On a Mac that usually means waiting for the app to ship a new version, or cloning llama.cpp, building it with Metal yourself and pointing your tools at the new binary. Off Grid AI Desktop removes that step. It downloads newer Metal builds of llama.cpp, stable-diffusion.cpp and whisper.cpp in the background, checks that each one starts on your Mac, and uses it the next time a model loads.

To show how far this moves: the 0.0.55 Mac app ships with llama.cpp b11819. On 31 October 2026 the macOS engine index already offered b12000, so a Mac on 0.0.55 picks that up without downloading a new DMG.

What you need

A Mac with Apple silicon (M1 or later) running macOS 13 Ventura or newer. The engines inside the app are built for macOS 13 and up, and the app will not open on anything older.
Off Grid AI Desktop 0.0.55 or later, the current stable release. Download Off Grid AI Desktop for Mac.
An internet connection when an update downloads. Chat, images and transcription run offline once the engines and models are on disk.
No account and no payment. Engine updates are part of the free app.

There is nothing extra to install for the GPU. Every Mac engine is built with Metal, so it runs on the Apple silicon GPU from the first launch.

Which engines update on your Mac

Each line below is a separate engine with its own version. Here is what 0.0.55 ships with, next to what the macOS index offered on 31 October 2026:

Engine	What it runs	In the 0.0.55 app	In the index
llama.cpp	Chat and vision models	b11819	b12000
Prism (a llama.cpp fork)	Bonsai models	prism-b10787	prism-b10831
stable-diffusion.cpp	Image generation	master-987	master-1028
whisper.cpp	Transcription	v1.9.5	v1.9.5

The full Mac set came to about 360 MB in that index: roughly 121 MB for llama.cpp, 92 MB for Prism, 101 MB for stable-diffusion.cpp and 48 MB for whisper.cpp. The app only downloads the files it does not already have, so a later update is usually one engine, not all four.

A nightly job moves each engine to its newest upstream release, builds it, checks it and publishes it. A new llama.cpp tag reaches your Mac without anyone shipping a new version of the app.

Step 1: Open the Engines card

Open Off Grid AI Desktop and click Settings in the sidebar.
Find the Engines card. Its summary reads "Keep chat, image and transcription engines current, and manage GPU components."
The first line in the card reads "Metal support is included on macOS. No extra GPU download is needed." Below it is the Keep engines up to date switch.

Step 2: Leave Keep engines up to date on

The switch is on by default. With it on, the app checks the engine index 1 minute after launch and then every 6 hours. For each engine it has an older build of, it:

Downloads only the files it does not already have.
Checks that the index carries a valid signature, then checks each file's SHA-256 hash.
Starts the new engine once, with a 30-second limit, to prove it runs on your Mac.
Makes it the first engine the app looks for.

The line under the switch tells you where it is: "Checking for newer engines...", a download count such as "Downloading 40 MB of 121 MB", "Checking the new engine starts...", and then "Engines are up to date. A new engine is used the next time a model loads."

Under that line, the app lists the version of each engine it installed, for example Chat: b12000 · Bonsai: prism-b10831-9d9efea · Images: master-1028-37bc88b · Transcription: v1.9.5. That answers "which llama.cpp am I running" without opening Terminal.

Step 3: Check for an update now

You do not have to wait 6 hours. Under the version list, click the round-arrow button, Check for engine updates. It spins while the check runs and stays disabled until it finishes.

When the status reads "Engines are up to date", reload your chat model so it starts on the new engine: switch to another model and back, or quit the app with Cmd+Q and open it again. A model that is already running keeps the engine it started with until then, so nothing changes in the middle of an answer.

Step 4: Load the model that needed it

Once the new llama.cpp is in, load the model that refused before. If it failed because its architecture was newer than your engine, it now loads like any other GGUF. You did not download a new DMG, and the app did not restart on its own.

If you keep models loaded in the Gateway

If you keep several chat models in memory through Gateway > Keep loaded, a new llama.cpp does not restart them while they answer requests from your editor or scripts. A card appears instead, "Engine update ready for kept models", with a Restart models button. They keep serving on the old build until you click it, so you choose when the short reload happens.

When an update fails

If a download breaks off or a new engine does not start, the status line turns red and says why, for example "The new llama engine did not start" followed by the error. The app keeps using the engine it had and tries again at the next check.

Three layers keep you running. The new build only becomes active after it passes its start test. The previous build stays on disk behind it. The engines that shipped inside the app stay behind both. A bad build cannot leave your Mac without a working engine.

If you would rather stay on the engines your app release shipped with, turn Keep engines up to date off. The status line then reads "Engines stay on the version this app shipped with."

FAQ

How do I update llama.cpp in Off Grid AI Desktop on my Mac?
You do not need to do anything. With Keep engines up to date on, the app installs newer Metal builds in the background. To check at once, open Settings > Engines and click Check for engine updates.

How do I know which llama.cpp version my Mac is running?
Open Settings > Engines. The line under the status lists each installed engine and its version, such as Chat: b12000. If the line is empty, the app is still on the engines it shipped with.

Do I need a new DMG to run a brand-new model?
Usually not. If the model needs a newer llama.cpp than your app shipped with, the engine update brings it. Reload the model once the status reads "Engines are up to date."

Does the updated llama.cpp still use Metal on Apple silicon?
Yes. Every macOS engine in the index is built with Metal, the same as the ones inside the app. There is no separate GPU download on a Mac.

Is it safe to let the app install new engine builds?
Each update comes from a signed index, and every file is checked against its SHA-256 hash before use. A new engine must start in a 30-second test before the app uses it, and the previous build stays on disk as a fallback.

Can I stop engines from updating?
Yes. Turn off Keep engines up to date in Settings > Engines. The engines stay on the version this app release shipped with.