From ba882a0425f43246de1167eb476d3aa4f2e55302 Mon Sep 17 00:00:00 2001
From: localai-org-maint-bot
 <306269227+localai-org-maint-bot@users.noreply.github.com>
Date: Sun, 25 Oct 2026 09:53:06 +0000
Subject: [PATCH] feat(gallery): add official Flint INT4 variant

Offer the publisher Q4_1 export alongside the existing Flint builds.
Pin the weights and document explicit variant installation.

Assisted-by: Codex:gpt-6

(cherry picked from commit 819e49482ed49a4d2983611d71ac71e7ffbf3197, mudler/LocalAI#12862)
Assisted-by: Claude:claude-sonnet-5-5
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
---
 gallery/index.yaml        | 42 ++++++++++++++++++++++++++++++++
 gallery/models/index.yaml | 50 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 92 insertions(+)

diff --git a/gallery/index.yaml b/gallery/index.yaml
index 97f2d048a..d475dd4c1 100644
--- a/gallery/index.yaml
+++ b/gallery/index.yaml
@@ -79990,6 +79990,7 @@
   variants:
     - model: flint-q2
     - model: flint-q6
+    - model: flint-int4-q4-1
 - url: github:mudler/LocalAI/gallery/virtual.yaml@master
   name: flint-q2
   description: |
@@ -80066,6 +80067,47 @@
       top_p: 0.95
     template:
       use_tokenizer_template: true
+- url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  name: flint-int4-q4-1
+  description: |
+    Flint is a coding-agent mixture-of-experts model from Recursive, with
+    39.8B total and about 2.5B active parameters per token. This official
+    Q4_1 GGUF preserves the trained INT4 codes from the MLX export, with
+    F32 floating tensors. The 26 GB file includes the tokenizer and chat
+    template. This entry uses a 32,768-token default context.
+  license: apache-2.0
+  urls:
+    - https://huggingface.co/RecursiveSI/flint
+    - https://huggingface.co/RecursiveSI/flint-int4-gguf
+  tags:
+    - llm
+    - gguf
+    - cpu
+    - gpu
+    - moe
+    - coding
+    - agent
+    - reasoning
+    - tools
+  files:
+    - filename: llama-cpp/models/RecursiveSI/flint-int4-gguf/flint-int4-q4_1.gguf
+      sha256: 43c07baccdba42eb7528a9630ee15699b1942c297291dc1b39a061f508ee1a4b
+      uri: https://huggingface.co/RecursiveSI/flint-int4-gguf/resolve/9c47151036bc2c34331142dda04488bceab2bc9f/flint-int4-q4_1.gguf
+  overrides:
+    backend: llama-cpp
+    context_size: 32768
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      min_p: 0
+      model: llama-cpp/models/RecursiveSI/flint-int4-gguf/flint-int4-q4_1.gguf
+      temperature: 1
+      top_k: 20
+      top_p: 0.95
+    template:
+      use_tokenizer_template: true
 - url: github:mudler/LocalAI/gallery/virtual.yaml@master
   name: hy-memory-30b-a3b-q4
   description: |
diff --git a/gallery/models/index.yaml b/gallery/models/index.yaml
index 76b9ccc47..82b1b0c1b 100644
--- a/gallery/models/index.yaml
+++ b/gallery/models/index.yaml
@@ -98227,6 +98227,7 @@
     variants:
       - model: flint-q2
       - model: flint-q6
+      - model: flint-int4-q4-1
 - kind: model
   id: flint-q2
   name: flint-q2
@@ -98319,6 +98320,55 @@
         top_p: 0.95
       template:
         use_tokenizer_template: true
+- kind: model
+  id: flint-int4-q4-1
+  name: flint-int4-q4-1
+  version: 0.1.0
+  description: |
+    Flint is a coding-agent mixture-of-experts model from Recursive, with
+    39.8B total and about 2.5B active parameters per token. This official
+    Q4_1 GGUF preserves the trained INT4 codes from the MLX export, with
+    F32 floating tensors. The 26 GB file includes the tokenizer and chat
+    template. This entry uses a 32,768-token default context.
+  license: apache-2.0
+  urls:
+    - https://huggingface.co/RecursiveSI/flint
+    - https://huggingface.co/RecursiveSI/flint-int4-gguf
+  tags:
+    - llm
+    - gguf
+    - cpu
+    - gpu
+    - moe
+    - coding
+    - agent
+    - reasoning
+    - tools
+  channel: gallery
+  policy:
+    requireSignature: false
+  schemaVersion: 1
+  model:
+    url: github:mudler/LocalAI/gallery/virtual.yaml@master
+    files:
+      - filename: llama-cpp/models/RecursiveSI/flint-int4-gguf/flint-int4-q4_1.gguf
+        sha256: 43c07baccdba42eb7528a9630ee15699b1942c297291dc1b39a061f508ee1a4b
+        uri: https://huggingface.co/RecursiveSI/flint-int4-gguf/resolve/9c47151036bc2c34331142dda04488bceab2bc9f/flint-int4-q4_1.gguf
+    overrides:
+      backend: llama-cpp
+      context_size: 32768
+      known_usecases:
+        - chat
+      options:
+        - use_jinja:true
+      parameters:
+        model: llama-cpp/models/RecursiveSI/flint-int4-gguf/flint-int4-q4_1.gguf
+        temperature: 1
+        top_k: 20
+        top_p: 0.95
+        min_p: 0
+      template:
+        use_tokenizer_template: true
 - kind: model
   id: hy-memory-30b-a3b-q4
   name: hy-memory-30b-a3b-q4