From 1c17d6fe9df1918e8f9a49cc16eca541a7c0438d Mon Sep 17 00:00:00 2001
From: localai-org-maint-bot
 <306269227+localai-org-maint-bot@users.noreply.github.com>
Date: Wed, 7 Oct 2026 08:06:09 +0000
Subject: [PATCH] chore(gallery): add FrogNano Q5 and Q6 variants

Fill the quantization gap between the existing Q4 and Q8 builds. Pin both downloads to the same publisher revision and verify SHA256 values against LFS metadata and x-linked-etag headers.

Assisted-by: Codex:gpt-6
---
 docs/content/features/model-gallery.md | 15 +++++
 gallery/index.yaml                     | 80 ++++++++++++++++++++++++++
 2 files changed, 95 insertions(+)

diff --git a/docs/content/features/model-gallery.md b/docs/content/features/model-gallery.md
index 3bdfc6d41..86c15d300 100644
--- a/docs/content/features/model-gallery.md
+++ b/docs/content/features/model-gallery.md
@@ -23,6 +23,21 @@ GPT and text generation models might have a license which is not permissive for
 
 ![output](https://github.com/mudler/LocalAI/assets/2420543/7b16676e-d5b1-4c97-89bd-9fa5065c21ad)
 
+## FrogNano coding model
+
+`frognano-4b-2609` offers Q4_K_M, Q5_K_M, Q6_K, and Q8_0 GGUF builds
+through llama.cpp. The Q5 and Q6 variants provide intermediate sizes between
+the Q4 and Q8 builds. To select one explicitly:
+
+```bash
+local-ai models install frognano-4b-2609 --variant frognano-4b-2609-q5
+local-ai models install frognano-4b-2609 --variant frognano-4b-2609-q6
+```
+
+All builds use the embedded chat template and a 32,768-token default context.
+They are configured for text-based coding and tool use. The publisher does
+not support the checkpoint's inherited image and video components.
+
 ## How it works
 
 Open **Models** in the WebUI. It is the canonical page for a model's complete
diff --git a/gallery/index.yaml b/gallery/index.yaml
index 31e6a6aa1..e978b48fb 100644
--- a/gallery/index.yaml
+++ b/gallery/index.yaml
@@ -69293,6 +69293,8 @@
 - name: frognano-4b-2609
   url: github:mudler/LocalAI/gallery/virtual.yaml@master
   variants:
+    - model: frognano-4b-2609-q5
+    - model: frognano-4b-2609-q6
     - model: frognano-4b-2609-q8
   urls:
     - https://huggingface.co/microsoft/FrogNano-4B-2609
@@ -69331,6 +69333,84 @@
     - filename: FrogNano-4B-2609-Q4_K_M.gguf
       uri: https://huggingface.co/bartowski/FrogNano-4B-2609-GGUF/resolve/dfbb801e04ddf045ffd245b9f1a546ed0c7d960f/FrogNano-4B-2609-Q4_K_M.gguf
       sha256: 01fba00c57e2976399ae4947f0b4fdde4623d3e410a435bf2ae39ddf164e9b0a
+- name: frognano-4b-2609-q5
+  url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  urls:
+    - https://huggingface.co/microsoft/FrogNano-4B-2609
+    - https://huggingface.co/bartowski/FrogNano-4B-2609-GGUF
+  description: |
+    FrogNano is Microsoft's Qwen3.5-4B derivative for repository-level coding
+    and tool use. This Q5_K_M GGUF build uses the embedded chat template
+    with a 32K-token default context. Configured for text input; upstream
+    does not evaluate or support its inherited image and video components.
+  license: mit
+  tags:
+    - llm
+    - gguf
+    - cpu
+    - gpu
+    - coding
+    - reasoning
+  last_checked: "2026-10-07"
+  overrides:
+    backend: llama-cpp
+    context_size: 32768
+    function:
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: FrogNano-4B-2609-Q5_K_M.gguf
+      temperature: 0.6
+      repeat_penalty: 1.0
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: FrogNano-4B-2609-Q5_K_M.gguf
+      uri: https://huggingface.co/bartowski/FrogNano-4B-2609-GGUF/resolve/dfbb801e04ddf045ffd245b9f1a546ed0c7d960f/FrogNano-4B-2609-Q5_K_M.gguf
+      sha256: f37ff83a6229a127866c1deb414ae36f6c9c30e1b21d8e0bbe313f826cb23601
+- name: frognano-4b-2609-q6
+  url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  urls:
+    - https://huggingface.co/microsoft/FrogNano-4B-2609
+    - https://huggingface.co/bartowski/FrogNano-4B-2609-GGUF
+  description: |
+    FrogNano is Microsoft's Qwen3.5-4B derivative for repository-level coding
+    and tool use. This Q6_K GGUF build uses the embedded chat template
+    with a 32K-token default context. Configured for text input; upstream
+    does not evaluate or support its inherited image and video components.
+  license: mit
+  tags:
+    - llm
+    - gguf
+    - cpu
+    - gpu
+    - coding
+    - reasoning
+  last_checked: "2026-10-07"
+  overrides:
+    backend: llama-cpp
+    context_size: 32768
+    function:
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: FrogNano-4B-2609-Q6_K.gguf
+      temperature: 0.6
+      repeat_penalty: 1.0
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: FrogNano-4B-2609-Q6_K.gguf
+      uri: https://huggingface.co/bartowski/FrogNano-4B-2609-GGUF/resolve/dfbb801e04ddf045ffd245b9f1a546ed0c7d960f/FrogNano-4B-2609-Q6_K.gguf
+      sha256: 855278f785dc23fc4c150bf74e20ca6ed53036c06fb04abfd0ddc60d02116b6a
 - name: frognano-4b-2609-q8
   url: github:mudler/LocalAI/gallery/virtual.yaml@master
   urls: