From 2fa431bf6f5fb4fd2858a4ed1fae4368167d40d7 Mon Sep 17 00:00:00 2001
From: localai-org-maint-bot <bot-opensource@localaisrl.com>
Date: Mon, 17 Aug 2026 08:57:13 +0200
Subject: [PATCH] feat(gallery): add Qwen3.8 9B variants (#11557)

Add Q4_K_M and Q8_0 llama.cpp builds for the text-only Qwen3.8 9B distillation.

Assisted-by: Codex:gpt-5

Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
---
 gallery/index.yaml | 85 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 85 insertions(+)

diff --git a/gallery/index.yaml b/gallery/index.yaml
index dfc5a1a0d..417182d3a 100644
--- a/gallery/index.yaml
+++ b/gallery/index.yaml
@@ -158,6 +158,91 @@
     - filename: llama-cpp/mmproj/qwen3.8-27b/mmproj-Qwen3.8-27B-Q8_0.gguf
       uri: huggingface://ggml-org/Qwen3.8-27B-GGUF/mmproj-Qwen3.8-27B-Q8_0.gguf
       sha256: 2e968a6af97ce35d8971890b257b9b7edabf20ad91450501fa53162a19ee33eb
+- &qwen3-8-9b
+  name: "qwen3.8-9b-q4"
+  variants:
+    - model: qwen3.8-9b-q8
+  url: "github:mudler/LocalAI/gallery/virtual.yaml@master"
+  urls:
+    - https://huggingface.co/empero-ai/Qwen3.8-9B
+    - https://huggingface.co/empero-ai/Qwen3.8-9B-GGUF
+  description: |
+    Qwen3.8-9B is Empero AI's full-parameter distillation of Qwen3.8 2.4T
+    A95B into the dense Qwen3.5-9B architecture. It targets reasoning,
+    mathematics, coding, instruction following, and tool use, and supports a
+    native 262K-token context window. This default entry uses Q4_K_M weights;
+    a higher-quality Q8_0 build is available as a variant.
+  license: "apache-2.0"
+  tags:
+    - llm
+    - gguf
+    - cpu
+    - gpu
+    - qwen
+    - reasoning
+    - thinking
+    - coding
+    - agent
+    - tools
+    - long-context
+  icon: https://qianwen-res.oss-cn-beijing.aliyuncs.com/logo_qwen.jpg
+  last_checked: "2026-08-17"
+  overrides:
+    backend: llama-cpp
+    context_size: 262144
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      min_p: 0
+      model: llama-cpp/models/qwen3.8-9b/Qwen3.8-9B-Q4_K_M.gguf
+      presence_penalty: 0
+      repeat_penalty: 1
+      temperature: 0.6
+      top_k: 20
+      top_p: 0.95
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/qwen3.8-9b/Qwen3.8-9B-Q4_K_M.gguf
+      uri: huggingface://empero-ai/Qwen3.8-9B-GGUF/Qwen3.8-9B-Q4_K_M.gguf
+      sha256: df13d66021cef676f82be74053220fd75af6bf2a6a7fb77f5222ab9e50744a7a
+- !!merge <<: *qwen3-8-9b
+  name: "qwen3.8-9b-q8"
+  variants: []
+  description: |
+    Qwen3.8-9B in the higher-quality Q8_0 GGUF format. This variant preserves
+    more model fidelity for hosts with enough memory.
+  overrides:
+    backend: llama-cpp
+    context_size: 262144
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      min_p: 0
+      model: llama-cpp/models/qwen3.8-9b/Qwen3.8-9B-Q8_0.gguf
+      presence_penalty: 0
+      repeat_penalty: 1
+      temperature: 0.6
+      top_k: 20
+      top_p: 0.95
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/qwen3.8-9b/Qwen3.8-9B-Q8_0.gguf
+      uri: huggingface://empero-ai/Qwen3.8-9B-GGUF/Qwen3.8-9B-Q8_0.gguf
+      sha256: 79ca5d342a07922f2bbf38c8d892a79a3c8620c65feaf4b1c66b7830ae724db8
 - &twil-lm3
   name: "twil-lm3-q4"
   variants: