From 3b73ecd1e294607450334d9027ef2cff06c4b8bd Mon Sep 17 00:00:00 2001
From: localai-org-maint-bot
 <306269227+localai-org-maint-bot@users.noreply.github.com>
Date: Fri, 9 Oct 2026 04:03:13 +0000
Subject: [PATCH] chore(gallery): add TwIL Pro Q5 and Q6 variants

Fill the gap between the existing Q4 and Q8 builds with the
publisher's Q5_K_M and Q6_K GGUF files. Preserve the runtime settings
and license notice, and document explicit variant installation.

Assisted-by: Codex:gpt-6
---
 docs/content/features/model-gallery.md | 25 +++++++++
 gallery/index.yaml                     | 72 ++++++++++++++++++++++++++
 2 files changed, 97 insertions(+)

diff --git a/docs/content/features/model-gallery.md b/docs/content/features/model-gallery.md
index a9e10330f..c07e9d6e6 100644
--- a/docs/content/features/model-gallery.md
+++ b/docs/content/features/model-gallery.md
@@ -133,6 +133,31 @@ frees. Removal then waits 30 seconds, during which you can undo it; the delete
 request is sent only when that time ends. If you leave the page during the wait,
 nothing is deleted.
 
+## TwIL-LM3-Pro
+
+TwIL-LM3-Pro is a 3.66B Granite 4.2 fine-tune for formal logic, entailment,
+semantic parsing, and Lean formalization. The gallery offers Q4_K_M, Q5_K_M,
+Q6_K, and Q8_0 GGUF builds for llama.cpp, with a 32,768-token context and
+its embedded reasoning template. Q5 and Q6 provide options between Q4 and Q8
+in download size and quantization precision.
+
+Install with automatic variant selection:
+
+```bash
+local-ai models install twil-lm3-pro-q4
+```
+
+To select Q6_K explicitly:
+
+```bash
+local-ai models install twil-lm3-pro-q4 --variant twil-lm3-pro-q6
+```
+
+Use `--variant twil-lm3-pro-q5` for Q5_K_M. The
+[model license](https://huggingface.co/webAI-Official/TwIL-LM3-Pro/blob/main/LICENSE.md)
+permits noncommercial research and educational use only. See the
+[model card](https://huggingface.co/webAI-Official/TwIL-LM3-Pro) for task examples.
+
 ## Cyber-Ornith 1.5 9B
 
 Cyber-Ornith 1.5 is a Qwen3.5 fine-tune for security auditing, terminal tasks, and tool use.
diff --git a/gallery/index.yaml b/gallery/index.yaml
index 17a7cdfa1..63ac8c4e4 100644
--- a/gallery/index.yaml
+++ b/gallery/index.yaml
@@ -68100,6 +68100,8 @@
 - name: twil-lm3-pro-q4
   variants:
     - model: twil-lm3-pro-q8
+    - model: twil-lm3-pro-q5
+    - model: twil-lm3-pro-q6
   url: github:mudler/LocalAI/gallery/virtual.yaml@master
   urls:
     - https://huggingface.co/ibm-granite/granite-4.2-3b
@@ -68169,6 +68171,76 @@
     - filename: llama-cpp/models/twil-lm3-pro/TwIL-LM3-Pro-Q8_0.gguf
       uri: https://huggingface.co/webAI-Official/TwIL-LM3-Pro/resolve/4214aaa4020cd51ac418dc6c7243c076665a811d/TwIL-LM3-Pro-Q8_0.gguf
       sha256: 4b4ba4a94bc5cbe4783f43be3ae428be9b5d9e3dffb725c36ae4908d27067906
+- name: twil-lm3-pro-q5
+  url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  urls:
+    - https://huggingface.co/ibm-granite/granite-4.2-3b
+    - https://huggingface.co/webAI-Official/TwIL-LM3-Pro
+  description: |
+    TwIL-LM3-Pro is a 3.66B Granite 4.2 fine-tune for formal logic,
+    entailment, semantic parsing, and Lean formalization. This Q5_K_M GGUF
+    uses llama.cpp with the embedded reasoning template and greedy sampling.
+    The model license permits noncommercial research and educational use only.
+  license: webai-non-commercial-license-ver.-1.0
+  tags:
+    - llm
+    - gguf
+    - cpu
+    - gpu
+    - reasoning
+    - formal-logic
+  last_checked: "2026-10-09"
+  overrides:
+    backend: llama-cpp
+    context_size: 32768
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/twil-lm3-pro/TwIL-LM3-Pro-Q5_K_M.gguf
+      temperature: 0
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/twil-lm3-pro/TwIL-LM3-Pro-Q5_K_M.gguf
+      uri: https://huggingface.co/webAI-Official/TwIL-LM3-Pro/resolve/4214aaa4020cd51ac418dc6c7243c076665a811d/TwIL-LM3-Pro-Q5_K_M.gguf
+      sha256: 7c367c005c5fb8af641434165220d476352704860bd1f66fd61dd0cf56936d4d
+- name: twil-lm3-pro-q6
+  url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  urls:
+    - https://huggingface.co/ibm-granite/granite-4.2-3b
+    - https://huggingface.co/webAI-Official/TwIL-LM3-Pro
+  description: |
+    TwIL-LM3-Pro is a 3.66B Granite 4.2 fine-tune for formal logic,
+    entailment, semantic parsing, and Lean formalization. This Q6_K GGUF
+    uses llama.cpp with the embedded reasoning template and greedy sampling.
+    The model license permits noncommercial research and educational use only.
+  license: webai-non-commercial-license-ver.-1.0
+  tags:
+    - llm
+    - gguf
+    - cpu
+    - gpu
+    - reasoning
+    - formal-logic
+  last_checked: "2026-10-09"
+  overrides:
+    backend: llama-cpp
+    context_size: 32768
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/twil-lm3-pro/TwIL-LM3-Pro-Q6_K.gguf
+      temperature: 0
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/twil-lm3-pro/TwIL-LM3-Pro-Q6_K.gguf
+      uri: https://huggingface.co/webAI-Official/TwIL-LM3-Pro/resolve/4214aaa4020cd51ac418dc6c7243c076665a811d/TwIL-LM3-Pro-Q6_K.gguf
+      sha256: 0d1c2333e2bc60840f1e6df63cef935b0d1d0434e98c8b5f0e4fb20f2c3a6a63
 - name: gemma-4-e4b-obliterated-q4
   variants:
     - model: gemma-4-e4b-obliterated-q8