From 804dc109681b2fe467b69ee187a602e82decc0d4 Mon Sep 17 00:00:00 2001
From: localai-org-maint-bot <bot-opensource@localaisrl.com>
Date: Sun, 16 Aug 2026 21:44:22 +0200
Subject: [PATCH] feat(gallery): add Tess 4 27B variants (#11547)

Add Q4_K_M and Q8_0 multimodal builds. Include an MTP-enabled Q4_K_M build for speculative decoding.

Assisted-by: Codex:gpt-5

Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com>
---
 gallery/index.yaml | 134 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 134 insertions(+)

diff --git a/gallery/index.yaml b/gallery/index.yaml
index e20ea1a2b..44c6fd0f8 100644
--- a/gallery/index.yaml
+++ b/gallery/index.yaml
@@ -2370,6 +2370,140 @@
     - filename: llama-cpp/mmproj/Qwythos-9B-v2-MTP-Q4_K_M/mmproj-Qwythos-9B-v2-BF16.gguf
       sha256: 0d1687cb33124c78acab788b342d4a2eaf85b3035e87c3abe4ee9d0b84ddb4f5
       uri: https://huggingface.co/empero-ai/Qwythos-9B-v2-GGUF/resolve/main/mmproj-Qwythos-9B-v2-BF16.gguf
+- &tess-4-27b
+  name: "tess-4-27b"
+  variants:
+    - model: tess-4-27b-q8
+    - model: tess-4-27b-mtp
+  url: "github:mudler/LocalAI/gallery/virtual.yaml@master"
+  urls:
+    - https://huggingface.co/migtissera/Tess-4-27B
+    - https://huggingface.co/migtissera/Tess-4-27B-GGUF
+  description: |
+    Tess-4-27B is an Apache-2.0 agentic and reasoning model built on
+    Qwen3.6-27B. It scales its thinking depth to the task and supports tool
+    use, long-context work, and image input. This default entry uses the
+    Q4_K_M GGUF quantization and the shared F16 vision projector.
+  license: "apache-2.0"
+  tags:
+    - llm
+    - gguf
+    - cpu
+    - gpu
+    - qwen
+    - reasoning
+    - thinking
+    - agent
+    - tools
+    - long-context
+    - vision
+    - multimodal
+  last_checked: "2026-08-16"
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+      - vision
+    mmproj: llama-cpp/mmproj/tess-4-27b/mmproj-Tess-4-27B-F16.gguf
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/tess-4-27b/Tess-4-27B-Q4_K_M.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/tess-4-27b/Tess-4-27B-Q4_K_M.gguf
+      sha256: 99b94644e6b31dff920ec5eb1b8598edc95eed7fcc8c3058a98f42319fb2fef4
+      uri: huggingface://migtissera/Tess-4-27B-GGUF/Tess-4-27B-Q4_K_M.gguf
+    - filename: llama-cpp/mmproj/tess-4-27b/mmproj-Tess-4-27B-F16.gguf
+      sha256: 6be9dd0b2298225381277c2b7df9c4bbcc5b9d95b43b7299123f8392527cde69
+      uri: huggingface://migtissera/Tess-4-27B-GGUF/mmproj-Tess-4-27B-F16.gguf
+- !!merge <<: *tess-4-27b
+  name: "tess-4-27b-q8"
+  variants: []
+  description: |
+    Tess-4-27B is an Apache-2.0 agentic and reasoning model built on
+    Qwen3.6-27B. This entry uses the near-lossless Q8_0 GGUF quantization and
+    the shared F16 vision projector.
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+      - vision
+    mmproj: llama-cpp/mmproj/tess-4-27b-q8/mmproj-Tess-4-27B-F16.gguf
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/tess-4-27b-q8/Tess-4-27B-Q8_0.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/tess-4-27b-q8/Tess-4-27B-Q8_0.gguf
+      sha256: 35d0b640edc1a9c00f8089bada11c4a58fc825d9d0fbf4ac44f4054ec3cb3a57
+      uri: huggingface://migtissera/Tess-4-27B-GGUF/Tess-4-27B-Q8_0.gguf
+    - filename: llama-cpp/mmproj/tess-4-27b-q8/mmproj-Tess-4-27B-F16.gguf
+      sha256: 6be9dd0b2298225381277c2b7df9c4bbcc5b9d95b43b7299123f8392527cde69
+      uri: huggingface://migtissera/Tess-4-27B-GGUF/mmproj-Tess-4-27B-F16.gguf
+- !!merge <<: *tess-4-27b
+  name: "tess-4-27b-mtp"
+  variants: []
+  description: |
+    Tess-4-27B with its Q4_K_M multi-token prediction draft enabled for
+    speculative decoding. The main model verifies every proposed token, and
+    the entry also includes the shared F16 vision projector.
+  tags:
+    - llm
+    - gguf
+    - cpu
+    - gpu
+    - qwen
+    - reasoning
+    - thinking
+    - agent
+    - tools
+    - long-context
+    - vision
+    - multimodal
+    - mtp
+    - speculative
+  overrides:
+    backend: llama-cpp
+    draft_model: llama-cpp/models/tess-4-27b-mtp/mtp-Tess-4-27B-Q4_K_M.gguf
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+      - vision
+    mmproj: llama-cpp/mmproj/tess-4-27b-mtp/mmproj-Tess-4-27B-F16.gguf
+    options:
+      - use_jinja:true
+      - spec_type:draft-mtp
+      - spec_n_max:6
+      - spec_p_min:0.75
+    parameters:
+      model: llama-cpp/models/tess-4-27b-mtp/Tess-4-27B-Q4_K_M.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/tess-4-27b-mtp/Tess-4-27B-Q4_K_M.gguf
+      sha256: 99b94644e6b31dff920ec5eb1b8598edc95eed7fcc8c3058a98f42319fb2fef4
+      uri: huggingface://migtissera/Tess-4-27B-GGUF/Tess-4-27B-Q4_K_M.gguf
+    - filename: llama-cpp/models/tess-4-27b-mtp/mtp-Tess-4-27B-Q4_K_M.gguf
+      sha256: c8dd872f71f584eef1693779c98d9c6c81daaa9bf97a7d9b24ae9cb968a30a77
+      uri: huggingface://migtissera/Tess-4-27B-GGUF/mtp-Tess-4-27B-Q4_K_M.gguf
+    - filename: llama-cpp/mmproj/tess-4-27b-mtp/mmproj-Tess-4-27B-F16.gguf
+      sha256: 6be9dd0b2298225381277c2b7df9c4bbcc5b9d95b43b7299123f8392527cde69
+      uri: huggingface://migtissera/Tess-4-27B-GGUF/mmproj-Tess-4-27B-F16.gguf
 - &qwen3-6-14b-a3b-fablevibes
   name: "qwen3.6-14b-a3b-fablevibes"
   variants: