From c03dbd0a51b10509665d4d1d38c68e757ee244f3 Mon Sep 17 00:00:00 2001 From: "K. Isom" Date: Wed, 16 Sep 2026 00:40:47 -0700 Subject: [PATCH] straylight: add Qwen 3.8 27B uncensored to llama-server huihui Q4_K_L of Qwen3.8-27B (orcarouter GGUF is gated). Dense 27B coding agent, loaded on demand like Ornith. --- hw/straylight/default.nix | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/hw/straylight/default.nix b/hw/straylight/default.nix index eed629e..14b4fb7 100644 --- a/hw/straylight/default.nix +++ b/hw/straylight/default.nix @@ -49,6 +49,13 @@ let [ornith-1.0-35b] hf-repo = deepreinforce-ai/Ornith-1.0-35B-GGUF:Q4_K_M dedup-cache-models = true + + ; Qwen 3.8 27B dense, refusal-ablated. Preferred uncensored coding agent. + ; orcarouter/Qwen3.8-27B-Uncensored-GGUF is gated; huihui Q4_K_L keeps + ; ablation-target tensors at Q8_0. llama-server fetches on first load. + [qwen3.8-27b-uncensored] + hf-repo = huihui-ai/Huihui-Qwen3.8-27B-abliterated-GGUF:Q4_K_L + dedup-cache-models = true ''; # `llama-models` shows what the router has resident; `llama-unload` frees every loaded # model (e.g. before a gaming session), or just the ones named on the command line.