From 81f6898b69e466d9403f29255f907fe2d666bd69 Mon Sep 17 00:00:00 2001 From: Ettore Di Giacinto Date: Fri, 11 Sep 2026 22:37:52 +0000 Subject: [PATCH] gallery: apply PR #11960 Assisted-by: localai-org-maint-bot:glm5.2 [gh] --- docs/content/features/model-gallery.md | 8 ++- gallery/index.yaml | 98 ++++++++++++++++++++++++++ 2 files changed, 104 insertions(+), 2 deletions(-) diff --git a/docs/content/features/model-gallery.md b/docs/content/features/model-gallery.md index 8f194d4ea..4cd6f719a 100644 --- a/docs/content/features/model-gallery.md +++ b/docs/content/features/model-gallery.md @@ -257,12 +257,16 @@ variant name for ordinary decoding without a draft model. ### Model variants -Ornith 1.5 9B offers Q4_K_M, Q5_K_M, Q6_K, and Q8_0 GGUF builds with a shared BF16 vision projector. To select Q6_K explicitly: +The `qwen3.5-9b-defiant-fable-mtp` entry offers Q4_K_M and Q8_0 builds for +llama.cpp, with and without multi-token prediction (MTP). Each build includes +the BF16 vision projector. To select the Q8_0 MTP build explicitly: ```bash -local-ai models install ornith-1.5-9b-q4 --variant ornith-1.5-9b-q6 +local-ai models install localai@qwen3.5-9b-defiant-fable-mtp --variant qwen3.5-9b-defiant-fable-q8-mtp ``` +Use `--variant qwen3.5-9b-defiant-fable-q8` for Q8_0 with ordinary decoding. + Some gallery entries offer several builds of the same model: different quantizations, or the same weights served by a different engine. Such an entry carries a `variants` list, and installing it normally lets LocalAI choose: diff --git a/gallery/index.yaml b/gallery/index.yaml index ea867b7bd..437b555b9 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -5117,6 +5117,8 @@ name: "qwen3.5-9b-defiant-fable-mtp" variants: - model: qwen3.5-9b-defiant-fable + - model: qwen3.5-9b-defiant-fable-q8-mtp + - model: qwen3.5-9b-defiant-fable-q8 url: "github:mudler/LocalAI/gallery/virtual.yaml@master" urls: - https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF @@ -5210,6 +5212,102 @@ - filename: llama-cpp/mmproj/qwen3.5-9b-defiant-fable/mmproj-BF16.gguf uri: huggingface://DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF/mmproj-BF16.gguf sha256: 853698ce7aa6c7ba732478bad280240969ddf7b0fcbf93900046f63903a83383 +- !!merge <<: *qwen3-5-9b-defiant-fable + name: "qwen3.5-9b-defiant-fable-q8-mtp" + variants: [] + urls: + - https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP + - https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF + description: | + Qwen3.5 9B Defiant Fable in Q8_0 GGUF format for multimodal reasoning, + coding, and creative writing. Includes the matching BF16 vision projector. + Enables multi-token prediction with the MTP weights. + tags: + - llm + - gguf + - cpu + - gpu + - qwen3.5 + - reasoning + - coding + - creative-writing + - uncensored + - vision + - multimodal + - mtp + last_checked: "2026-09-10" + overrides: + backend: llama-cpp + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - vision + mmproj: llama-cpp/mmproj/qwen3.5-9b-defiant-fable/mmproj-BF16.gguf + options: + - use_jinja:true + - spec_type:draft-mtp + - spec_n_max:6 + - spec_p_min:0.75 + parameters: + model: llama-cpp/models/qwen3.5-9b-defiant-fable/Qwen3.5-9B-The-Defiant-Fable-Uncnr-Heretic-NEO-MAX-MTP-Q8_0.gguf + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/qwen3.5-9b-defiant-fable/Qwen3.5-9B-The-Defiant-Fable-Uncnr-Heretic-NEO-MAX-MTP-Q8_0.gguf + uri: https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF/resolve/239d236baf9172e5a4663add74bd34fd297b1c08/Qwen3.5-9B-The-Defiant-Fable-Uncnr-Heretic-NEO-MAX-MTP-Q8_0.gguf + sha256: 1830f3beffb43850143bf45ee6f9d637f82ca22ddd9e08e40be1399062be32a2 + - filename: llama-cpp/mmproj/qwen3.5-9b-defiant-fable/mmproj-BF16.gguf + uri: https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF/resolve/239d236baf9172e5a4663add74bd34fd297b1c08/mmproj-BF16.gguf + sha256: 853698ce7aa6c7ba732478bad280240969ddf7b0fcbf93900046f63903a83383 +- !!merge <<: *qwen3-5-9b-defiant-fable + name: "qwen3.5-9b-defiant-fable-q8" + urls: + - https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP + - https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF + variants: [] + description: | + Qwen3.5 9B Defiant Fable in Q8_0 GGUF format for multimodal reasoning, + coding, and creative writing. Includes the matching BF16 vision projector. + Uses ordinary decoding without multi-token prediction. + tags: + - llm + - gguf + - cpu + - gpu + - qwen3.5 + - reasoning + - coding + - creative-writing + - uncensored + - vision + - multimodal + last_checked: "2026-09-10" + overrides: + backend: llama-cpp + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - vision + mmproj: llama-cpp/mmproj/qwen3.5-9b-defiant-fable/mmproj-BF16.gguf + options: + - use_jinja:true + parameters: + model: llama-cpp/models/qwen3.5-9b-defiant-fable/Qwen3.5-9B-The-Defiant-Fable-Uncnr-Heretic-NEO-MAX-Q8_0.gguf + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/qwen3.5-9b-defiant-fable/Qwen3.5-9B-The-Defiant-Fable-Uncnr-Heretic-NEO-MAX-Q8_0.gguf + uri: https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF/resolve/239d236baf9172e5a4663add74bd34fd297b1c08/Qwen3.5-9B-The-Defiant-Fable-Uncnr-Heretic-NEO-MAX-Q8_0.gguf + sha256: a0ed28f9af669f48ff2138f1f016bda3f1b3daf305cc3bb9c501f4c6b48e48a5 + - filename: llama-cpp/mmproj/qwen3.5-9b-defiant-fable/mmproj-BF16.gguf + uri: https://huggingface.co/DavidAU/Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP-GGUF/resolve/239d236baf9172e5a4663add74bd34fd297b1c08/mmproj-BF16.gguf + sha256: 853698ce7aa6c7ba732478bad280240969ddf7b0fcbf93900046f63903a83383 - &qwen3-8-27b-cold-fusion name: "qwen3.8-27b-cold-fusion-q4-mtp" variants: