From 44413a9d06bf5bc52ce088ba8ca74e5a2e8bee26 Mon Sep 17 00:00:00 2001 From: localai-org-maint-bot Date: Fri, 14 Aug 2026 19:02:32 +0200 Subject: [PATCH] feat(gallery): add Qwen3.8 27B variants (#11519) Add the official Q4_K_M and Q8_0 GGUF files with the shared vision projector. Include an MTP variant 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 | 159 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 159 insertions(+) diff --git a/gallery/index.yaml b/gallery/index.yaml index 28541bab8..40c9210b7 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -1,4 +1,163 @@ --- +- &qwen3-8-27b + name: "qwen3.8-27b-q4" + variants: + - model: qwen3.8-27b-q4-mtp + - model: qwen3.8-27b-q8 + url: "github:mudler/LocalAI/gallery/virtual.yaml@master" + urls: + - https://huggingface.co/Qwen/Qwen3.8-27B + - https://huggingface.co/ggml-org/Qwen3.8-27B-GGUF + description: | + Qwen3.8-27B is Qwen's dense 27B vision-language model for reasoning, + coding, tool use, and long-running agent tasks. It accepts text, images, + and video, and it supports a native context window of 262K tokens. + + This default entry uses the official Q4_K_M GGUF and Q8_0 vision + projector. The linked variants add MTP speculative decoding or use the + higher-quality Q8_0 model. + license: "apache-2.0" + tags: + - llm + - gguf + - cpu + - gpu + - qwen + - reasoning + - thinking + - coding + - agent + - tools + - vision + - multimodal + - long-context + icon: https://qianwen-res.oss-cn-beijing.aliyuncs.com/logo_qwen.jpg + last_checked: "2026-08-14" + overrides: + backend: llama-cpp + context_size: 262144 + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - vision + mmproj: llama-cpp/mmproj/qwen3.8-27b/mmproj-Qwen3.8-27B-Q8_0.gguf + options: + - use_jinja:true + parameters: + min_p: 0 + model: llama-cpp/models/qwen3.8-27b/Qwen3.8-27B-Q4_K_M.gguf + presence_penalty: 0 + repeat_penalty: 1 + temperature: 1 + top_k: 20 + top_p: 0.95 + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/qwen3.8-27b/Qwen3.8-27B-Q4_K_M.gguf + uri: huggingface://ggml-org/Qwen3.8-27B-GGUF/Qwen3.8-27B-Q4_K_M.gguf + sha256: 31629f53165ab6a7dad8c9847dcfd1fdf55829dac1e6e748f4a68581b0033d34 + - 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 +- !!merge <<: *qwen3-8-27b + name: "qwen3.8-27b-q4-mtp" + variants: [] + description: | + Qwen3.8-27B with the official Q4_K_M model and Q4_0 MTP draft model. + MTP speculative decoding can increase generation speed by proposing + multiple tokens for the target model to verify. + tags: + - llm + - gguf + - gpu + - qwen + - reasoning + - thinking + - coding + - agent + - tools + - vision + - multimodal + - long-context + - mtp + - speculative + overrides: + backend: llama-cpp + context_size: 262144 + draft_model: llama-cpp/models/qwen3.8-27b/mtp-Qwen3.8-27B-Q4_0.gguf + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + - vision + mmproj: llama-cpp/mmproj/qwen3.8-27b/mmproj-Qwen3.8-27B-Q8_0.gguf + options: + - use_jinja:true + - spec_type:draft-mtp + - spec_n_max:6 + - spec_p_min:0.75 + parameters: + min_p: 0 + model: llama-cpp/models/qwen3.8-27b/Qwen3.8-27B-Q4_K_M.gguf + presence_penalty: 0 + repeat_penalty: 1 + temperature: 1 + top_k: 20 + top_p: 0.95 + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/qwen3.8-27b/Qwen3.8-27B-Q4_K_M.gguf + uri: huggingface://ggml-org/Qwen3.8-27B-GGUF/Qwen3.8-27B-Q4_K_M.gguf + sha256: 31629f53165ab6a7dad8c9847dcfd1fdf55829dac1e6e748f4a68581b0033d34 + - filename: llama-cpp/models/qwen3.8-27b/mtp-Qwen3.8-27B-Q4_0.gguf + uri: huggingface://ggml-org/Qwen3.8-27B-GGUF/mtp-Qwen3.8-27B-Q4_0.gguf + sha256: 051a1764cff8c4f3ee6ae8b00593a0364c7539c67fa50ffc58f3f96509fca38e + - 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 +- !!merge <<: *qwen3-8-27b + name: "qwen3.8-27b-q8" + variants: [] + description: | + Qwen3.8-27B in the official Q8_0 GGUF format. This variant provides higher + 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 + - vision + mmproj: llama-cpp/mmproj/qwen3.8-27b/mmproj-Qwen3.8-27B-Q8_0.gguf + options: + - use_jinja:true + parameters: + min_p: 0 + model: llama-cpp/models/qwen3.8-27b/Qwen3.8-27B-Q8_0.gguf + presence_penalty: 0 + repeat_penalty: 1 + temperature: 1 + top_k: 20 + top_p: 0.95 + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/qwen3.8-27b/Qwen3.8-27B-Q8_0.gguf + uri: huggingface://ggml-org/Qwen3.8-27B-GGUF/Qwen3.8-27B-Q8_0.gguf + sha256: f5c702d8820d36fb55985bb238fc83ee3a313e920f4b752a437c3a6a9e14e4c8 + - 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 - &twil-lm3 name: "twil-lm3-q4" variants: