My model won't load

Most common cause: not enough free RAM.

Short answer

Most common cause: not enough free RAM. Second: the GGUF file is corrupted or incomplete. Third: unusual GGUF that Abliteration can’t parse. Check the Hardware card first, then re-download the file if RAM isn’t the issue.

Detailed explanation

Step 1 - Check RAM.

Open Settings → Hardware. Look at free RAM vs the model’s file size + ~2 GB overhead. If free RAM is less than that, the model literally won’t fit - Windows will refuse to allocate, or start swapping so hard the load times out.

Fix: close other apps (browser especially), then try loading again. Or drop to a smaller quant (Q4 instead of Q8) or smaller model.

Step 2 - Check file integrity.

If a download was interrupted or resumed incorrectly, the GGUF may be truncated or missing bytes. Signs:

  • Abliteration shows a parsing error on model add
  • Model was added but crashes on first load
  • File size on disk is smaller than what Hugging Face lists

Fix: re-download the file. Use a browser that resumes properly, or use wget -c / curl -C - in a terminal for reliable resume.

Step 3 - Unusual GGUF.

Rare, but happens with very new or custom-trained models. Abliteration parses the standard GGUF header format - non-standard extensions or old GGUF v1 files (pre-2024) may not parse.

Fix: find a mainstream re-quantization of the same model (search for the model name + “GGUF” on Hugging Face, prefer unsloth or bartowski uploads).

Step 4 - Disk space.

Rarely the cause but worth checking. Abliteration copies the GGUF into its models folder on add - if your disk is nearly full, the copy fails silently.

Fix: free up space on the drive holding %LOCALAPPDATA%, then try again.

Why it matters

“Won’t load” is the most opaque error state. Systematic elimination beats guessing: RAM → file → format → disk. Nine times out of ten it’s RAM.

In Abliteration specifically

If the fit-check dialog told you “won’t fit” on add, and you clicked Add anyway, then subsequent load failures are exactly what the dialog was warning about. Trust the dialog next time or plan to close apps before loading.

If Abliteration crashes when you try to open a model (not just fails to load), that’s a separate issue. Restart the app and try loading a different model to see if the problem is model-specific or app-wide.

Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration