From a3a910f833b41833b0bf83190b849720023519ac Mon Sep 17 00:00:00 2001 From: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com> Date: Mon, 12 Oct 2026 16:06:35 +0000 Subject: [PATCH] chore(gallery): add Turbo Brilliance 50X variants Add Q4_K_M and Q8_0 builds with the F16 vision projector for llama.cpp. Pin downloads and verify SHA256 against HF metadata and response headers. Document variant selection and keep the embedded chat template. Assisted-by: Codex:GPT-6 --- docs/content/features/model-gallery.md | 21 ++++++++ gallery/index.yaml | 72 ++++++++++++++++++++++++++ 2 files changed, 93 insertions(+) diff --git a/docs/content/features/model-gallery.md b/docs/content/features/model-gallery.md index a5af4a904..0ef902327 100644 --- a/docs/content/features/model-gallery.md +++ b/docs/content/features/model-gallery.md @@ -187,6 +187,27 @@ To select Q8 explicitly, run `local-ai models install cyber-tiel-coder-35b-a3b-q Both configurations use the embedded chat template and default to 32,768 context tokens. The [model card](https://huggingface.co/peculiar-ragdoll/Cyber-Tiel-Coder-35B-A3B-GGUF-MTP) describes its abliterated Ornith-1.5 base and MIT license. +## Qwen3.8-27B Turbo Brilliance 50X + +The gallery provides Q4_K_M and Q8_0 builds for llama.cpp. Both include the +F16 vision projector and use the model's embedded chat template. The default +context is 32,768 tokens. These entries do not enable MTP speculative decoding. + +Install with automatic variant selection: + +```bash +local-ai models install qwen3.8-27b-turbo-brilliance-50x-q4 +``` + +To select Q8_0 explicitly: + +```bash +local-ai models install --variant qwen3.8-27b-turbo-brilliance-50x-q8 qwen3.8-27b-turbo-brilliance-50x-q4 +``` + +See the [publisher's model card](https://huggingface.co/DavidAU/Qwen3.8-27B-Turbo-Brilliance-50X-Reasoning-Instruct-NEO-MAX-MTP-GGUF) +for reasoning modes, prompt syntax, and sampling guidance. + ## Qwen3.8-27B Agention Precision The gallery includes Agention Precision IQ4_XS and Q4_K_M GGUF builds of diff --git a/gallery/index.yaml b/gallery/index.yaml index 947b4be7d..7f97b077f 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -843,6 +843,78 @@ - filename: Hemmingway-1.Q8_0.gguf sha256: 10b566343db6905215c2d635285ffb43ada05a2f9b1e8e85562c8e366fac1809 uri: huggingface://mradermacher/Hemmingway-1-GGUF/Hemmingway-1.Q8_0.gguf +- name: qwen3.8-27b-turbo-brilliance-50x-q4 + variants: + - model: qwen3.8-27b-turbo-brilliance-50x-q8 + url: github:mudler/LocalAI/gallery/qwen3.yaml@master + urls: + - https://huggingface.co/Qwen/Qwen3.8-27B + - https://huggingface.co/DavidAU/Qwen3.8-27B-Turbo-Brilliance-50X-Reasoning-Instruct-NEO-MAX-MTP-GGUF + description: | + DavidAU's Turbo Brilliance 50X edition of Qwen3.8-27B includes a custom + chat template with selectable reasoning and instruction modes. This + Q4_K_M GGUF build includes the F16 vision projector for image input. + Uses the embedded template with a 32,768-token context by default. + license: apache-2.0 + tags: + - llm + - gguf + - cpu + - gpu + - reasoning + - vision + overrides: + context_size: 32768 + mmproj: qwen3.8-27b-turbo-brilliance-50x/mmproj-F16.gguf + parameters: + model: Qwen3.8-27B-Turbo-Brilliance-50X-RI-NEO-Q4_K_M.gguf + temperature: 0.7 + top_k: 64 + top_p: 0.95 + min_p: 0.05 + repeat_penalty: 1 + files: + - filename: Qwen3.8-27B-Turbo-Brilliance-50X-RI-NEO-Q4_K_M.gguf + sha256: 2e4fbc3b9c9cd45569dc506a87571a3dedfe213f9138cf4eece8d382c8d3794e + uri: https://huggingface.co/DavidAU/Qwen3.8-27B-Turbo-Brilliance-50X-Reasoning-Instruct-NEO-MAX-MTP-GGUF/resolve/35d2d177d8efe3acc102afb1fb96b0bc1fb0b5d9/Qwen3.8-27B-Turbo-Brilliance-50X-RI-NEO-Q4_K_M.gguf + - filename: qwen3.8-27b-turbo-brilliance-50x/mmproj-F16.gguf + sha256: cb0a71fe9b84af2552fc6260acf8e9773a29d115c683aba01c72a7ac537b212b + uri: https://huggingface.co/DavidAU/Qwen3.8-27B-Turbo-Brilliance-50X-Reasoning-Instruct-NEO-MAX-MTP-GGUF/resolve/35d2d177d8efe3acc102afb1fb96b0bc1fb0b5d9/mmproj-F16.gguf +- name: qwen3.8-27b-turbo-brilliance-50x-q8 + url: github:mudler/LocalAI/gallery/qwen3.yaml@master + urls: + - https://huggingface.co/Qwen/Qwen3.8-27B + - https://huggingface.co/DavidAU/Qwen3.8-27B-Turbo-Brilliance-50X-Reasoning-Instruct-NEO-MAX-MTP-GGUF + description: | + DavidAU's Turbo Brilliance 50X edition of Qwen3.8-27B includes a custom + chat template with selectable reasoning and instruction modes. This + Q8_0 GGUF build includes the F16 vision projector for image input. + Uses the embedded template with a 32,768-token context by default. + license: apache-2.0 + tags: + - llm + - gguf + - cpu + - gpu + - reasoning + - vision + overrides: + context_size: 32768 + mmproj: qwen3.8-27b-turbo-brilliance-50x/mmproj-F16.gguf + parameters: + model: Qwen3.8-27B-Turbo-Brilliance-50X-RI-NEO-Q8_0.gguf + temperature: 0.7 + top_k: 64 + top_p: 0.95 + min_p: 0.05 + repeat_penalty: 1 + files: + - filename: Qwen3.8-27B-Turbo-Brilliance-50X-RI-NEO-Q8_0.gguf + sha256: b354f6f683a55bdf17b5c51f0c203dc5c4da38bf877932fe82ff909811728f64 + uri: https://huggingface.co/DavidAU/Qwen3.8-27B-Turbo-Brilliance-50X-Reasoning-Instruct-NEO-MAX-MTP-GGUF/resolve/35d2d177d8efe3acc102afb1fb96b0bc1fb0b5d9/Qwen3.8-27B-Turbo-Brilliance-50X-RI-NEO-Q8_0.gguf + - filename: qwen3.8-27b-turbo-brilliance-50x/mmproj-F16.gguf + sha256: cb0a71fe9b84af2552fc6260acf8e9773a29d115c683aba01c72a7ac537b212b + uri: https://huggingface.co/DavidAU/Qwen3.8-27B-Turbo-Brilliance-50X-Reasoning-Instruct-NEO-MAX-MTP-GGUF/resolve/35d2d177d8efe3acc102afb1fb96b0bc1fb0b5d9/mmproj-F16.gguf - name: qwen3.8-27b url: github:mudler/LocalAI/gallery/virtual.yaml@master urls: