6 yesterday

"Unmasked" version of Llama3.2

tools
ollama run reecdev/llama3.2-unmasked

Applications

Claude Code
Claude Code ollama launch claude --model reecdev/llama3.2-unmasked
OpenClaw
OpenClaw ollama launch openclaw --model reecdev/llama3.2-unmasked
Hermes Agent
Hermes Agent ollama launch hermes --model reecdev/llama3.2-unmasked
Codex
Codex ollama launch codex --model reecdev/llama3.2-unmasked
OpenCode
OpenCode ollama launch opencode --model reecdev/llama3.2-unmasked

Models

View all →

Readme

Llama3.2-Unmasked

An “unmasked” version of Llama3.2.

What is unmasking?

Unmasking is a finetuning method similar to abliteration/jailbreaking that removes the original RLHF completely from a model instead of just removing refusals and then doing another finetuning step to make it conversational again.