From 0e3ee19ec4cdd0d33c2d4c60742836f491b66048 Mon Sep 17 00:00:00 2001 From: localai-org-maint-bot Date: Sat, 12 Sep 2026 23:08:23 +0200 Subject: [PATCH 1/2] fix(gallery): restore entries clobbered by batch PR applies (#12016) The batch of "gallery: apply PR" commits replayed gallery-agent diffs against a stale base. Each new top-of-file entry overwrote the entry above it instead of being inserted, which lost seven entries: - qwen3.8-27b-uncensored-q4/-q8 (#11705, overwritten by #11909) - qwen3.8-flash-next-uncensored (#11832, overwritten by #11841) - spark-x2.5-4b-q4/-q6/-q8 (#11923, overwritten by #11926) - deepseek-v4-flash-vision-exp (#11873): #11927 renamed its name line to qwopus3.8-27b-flash, which duplicated that entry and failed the "declares every entry name exactly once" gallery lint on master. Each restored entry is identical (YAML-equal) to the one in its PR head. Assisted-by: Claude:claude-opus-5 [Bash] Signed-off-by: Ettore Di Giacinto Co-authored-by: Ettore Di Giacinto --- gallery/index.yaml | 258 ++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 257 insertions(+), 1 deletion(-) diff --git a/gallery/index.yaml b/gallery/index.yaml index 5b0139e72..068db623a 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -46,7 +46,68 @@ - filename: llama-cpp/models/Spark-X2.5-4B-Q4_K_M/Spark-X2.5-4B-Q4_K_M.gguf sha256: adfcfa19a4ed6a5985da8bf565fe15f8e1a7e131d79bae2d19d48d1c40109428 uri: https://huggingface.co/XHToken/Spark-X2.5-4B-GGUF/resolve/main/Spark-X2.5-4B-Q4_K_M.gguf -- name: "qwopus3.8-27b-flash" +- name: "qwen3.8-flash-next-uncensored" + url: "github:mudler/LocalAI/gallery/virtual.yaml@master" + urls: + - https://huggingface.co/orcarouter/Qwen3.8-Flash-Next-Uncensored-GGUF + description: | + # Qwen3.8-Flash-Next + + > [!Note] + > This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers format. + > + > These artifacts are compatible with Hugging Face Transformers, vLLM, SGLang, TokenSpeed, etc. + + > [!Tip] + > For users seeking managed, scalable inference without infrastructure maintenance, the official Qwen API service is provided by Qwen Cloud. + > + > In particular, **Qwen3.8-Flash** is the official version based on Qwen3.8-Flash-Next with more production features, e.g., 1M context length by default, official built-in tools. For more information, please refer to the Qwen3.8-Flash Overview. + + As the frontier of foundation models pushes toward ever-larger parameter counts and ever-longer context windows, the question is no longer just how much we can scale, but how efficiently we can do so. Sustainable progress toward artificial general intelligence (AGI) that benefits everyone demands architectural innovation. Today, we are sharing a concrete step in that direction: Qwen3.8-Flash-Next. + + ... + license: "apache-2.0" + tags: + - llm + - gguf + - qwen + - reasoning + icon: https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3.8-Flash-Next/architecture.png + overrides: + backend: llama-cpp + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + mmproj: llama-cpp/mmproj/Qwen3.8-Flash-Next-Uncensored-Q4_K_M/mmproj-Qwen3.8-Flash-Next-Uncensored-F16.gguf + options: + - use_jinja:true + parameters: + min_p: 0 + model: llama-cpp/models/Qwen3.8-Flash-Next-Uncensored-Q4_K_M/Qwen3.8-Flash-Next-Uncensored-Q4_K_M-00001-of-00003.gguf + presence_penalty: 1.5 + repeat_penalty: 1 + temperature: 0.7 + top_k: 20 + top_p: 0.8 + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/Qwen3.8-Flash-Next-Uncensored-Q4_K_M/Qwen3.8-Flash-Next-Uncensored-Q4_K_M-00001-of-00003.gguf + sha256: '****************************************************************' + uri: https://huggingface.co/orcarouter/Qwen3.8-Flash-Next-Uncensored-GGUF/resolve/main/Qwen3.8-Flash-Next-Uncensored-Q4_K_M-00001-of-00003.gguf + - filename: llama-cpp/models/Qwen3.8-Flash-Next-Uncensored-Q4_K_M/Qwen3.8-Flash-Next-Uncensored-Q4_K_M-00002-of-00003.gguf + sha256: '****************************************************************' + uri: https://huggingface.co/orcarouter/Qwen3.8-Flash-Next-Uncensored-GGUF/resolve/main/Qwen3.8-Flash-Next-Uncensored-Q4_K_M-00002-of-00003.gguf + - filename: llama-cpp/models/Qwen3.8-Flash-Next-Uncensored-Q4_K_M/Qwen3.8-Flash-Next-Uncensored-Q4_K_M-00003-of-00003.gguf + sha256: '****************************************************************' + uri: https://huggingface.co/orcarouter/Qwen3.8-Flash-Next-Uncensored-GGUF/resolve/main/Qwen3.8-Flash-Next-Uncensored-Q4_K_M-00003-of-00003.gguf + - filename: llama-cpp/mmproj/Qwen3.8-Flash-Next-Uncensored-Q4_K_M/mmproj-Qwen3.8-Flash-Next-Uncensored-F16.gguf + sha256: '****************************************************************' + uri: https://huggingface.co/orcarouter/Qwen3.8-Flash-Next-Uncensored-GGUF/resolve/main/mmproj-Qwen3.8-Flash-Next-Uncensored-F16.gguf +- name: "deepseek-v4-flash-vision-exp" url: "github:mudler/LocalAI/gallery/virtual.yaml@master" urls: - https://huggingface.co/unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF @@ -2228,6 +2289,105 @@ - filename: llama-cpp/models/Qwen3.8-27B-DFlash2-Q4_K_M/Qwen3.8-27B-DFlash2-Q4_K_M.gguf uri: https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2-GGUF/resolve/main/Qwen3.8-27B-DFlash2-Q4_K_M.gguf sha256: 1a25c56858e1ebe93f2718ac1d49d1151f9323325c1bbfd6209370f4db131ebd +- &qwen3-8-27b-uncensored + name: "qwen3.8-27b-uncensored-q4" + variants: + - model: qwen3.8-27b-uncensored-q8 + url: "github:mudler/LocalAI/gallery/virtual.yaml@master" + urls: + - https://huggingface.co/Qwen/Qwen3.8-27B + - https://huggingface.co/JonathanColetti/Qwen3.8-27B-Uncensored-GGUF + description: | + Qwen3.8-27B-Uncensored reduces refusal behavior while retaining the base + model's text, vision, reasoning, and tool-use capabilities. Its integrated + MTP head supports speculative decoding without a separate draft model. + + This default entry uses the Q4_K_M GGUF and F16 vision projector. A + higher-quality Q8_0 model is available as a variant. + license: "apache-2.0" + tags: + - llm + - gguf + - cpu + - gpu + - qwen + - reasoning + - thinking + - tools + - vision + - multimodal + - long-context + - uncensored + - mtp + last_checked: "2026-08-24" + 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-uncensored/vision-f16.gguf + options: + - use_jinja:true + - spec_type:draft-mtp + - spec_n_max:3 + parameters: + min_p: 0 + model: llama-cpp/models/qwen3.8-27b-uncensored/Qwen3.8-27B-Uncensored-Q4_K_M.gguf + repeat_penalty: 1 + temperature: 0.6 + top_k: 20 + top_p: 0.95 + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/qwen3.8-27b-uncensored/Qwen3.8-27B-Uncensored-Q4_K_M.gguf + uri: huggingface://JonathanColetti/Qwen3.8-27B-Uncensored-GGUF/Qwen3.8-27B-Uncensored-Q4_K_M.gguf + sha256: 4c5e2db039e9325ac7724c8846c71356a24ad1cdfa28002d73ecb6be645f9675 + - filename: llama-cpp/mmproj/qwen3.8-27b-uncensored/vision-f16.gguf + uri: huggingface://JonathanColetti/Qwen3.8-27B-Uncensored-GGUF/Qwen3.8-27B-Uncensored-vision-f16.gguf + sha256: 5ac423f8a29059dc24e51bc6a43e9380dcd57a9347f28b62591e0b3f60b7081c +- !!merge <<: *qwen3-8-27b-uncensored + name: "qwen3.8-27b-uncensored-q8" + variants: [] + description: | + Qwen3.8-27B-Uncensored in the higher-quality Q8_0 GGUF format, with its + integrated MTP head and shared F16 vision projector. + 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-uncensored/vision-f16.gguf + options: + - use_jinja:true + - spec_type:draft-mtp + - spec_n_max:3 + parameters: + min_p: 0 + model: llama-cpp/models/qwen3.8-27b-uncensored/Qwen3.8-27B-Uncensored-Q8_0.gguf + repeat_penalty: 1 + temperature: 0.6 + top_k: 20 + top_p: 0.95 + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/qwen3.8-27b-uncensored/Qwen3.8-27B-Uncensored-Q8_0.gguf + uri: huggingface://JonathanColetti/Qwen3.8-27B-Uncensored-GGUF/Qwen3.8-27B-Uncensored-Q8_0.gguf + sha256: fb2cb9aa5407264aec6698faf41183f337809352abe0bd8fb750c0b689f4ced2 + - filename: llama-cpp/mmproj/qwen3.8-27b-uncensored/vision-f16.gguf + uri: huggingface://JonathanColetti/Qwen3.8-27B-Uncensored-GGUF/Qwen3.8-27B-Uncensored-vision-f16.gguf + sha256: 5ac423f8a29059dc24e51bc6a43e9380dcd57a9347f28b62591e0b3f60b7081c - &huihui-qwen3-8-27b name: "huihui-qwen3.8-27b-abliterated" variants: @@ -4535,6 +4695,102 @@ uri: huggingface://XHToken/Spark-X2.5-1.7B-GGUF/Spark-X2.5-1.7B-Q8_0.gguf sha256: cd77c03185a834bb1162a4b7713520be5838058bfc54873645beff470bb24442 +- &spark-x2-5-4b + name: "spark-x2.5-4b-q4" + url: "github:mudler/LocalAI/gallery/virtual.yaml@master" + variants: + - model: spark-x2.5-4b-q6 + - model: spark-x2.5-4b-q8 + urls: + - https://huggingface.co/XHToken/Spark-X2.5-4B + - https://huggingface.co/abenzerps/Spark-X2.5-4B-GGUF + description: | + Spark-X2.5-4B is XHToken's 4B text model for conversation, reasoning, + coding, and multilingual tasks. This entry uses Q4_K_M GGUF weights; + Q6_K and Q8_0 builds are available as variants. All builds use the + embedded Jinja chat template and a 32K-token default context. + license: "apache-2.0" + tags: + - llm + - gguf + - cpu + - gpu + - reasoning + - coding + - multilingual + last_checked: "2026-09-08" + overrides: + backend: llama-cpp + context_size: 32768 + known_usecases: + - chat + options: + - use_jinja:true + parameters: + model: llama-cpp/models/spark-x2.5-4b/Spark-X2.5-4B-Q4_K_M.gguf + temperature: 1 + top_p: 0.95 + top_k: -1 + min_p: 0 + repeat_penalty: 1 + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/spark-x2.5-4b/Spark-X2.5-4B-Q4_K_M.gguf + uri: huggingface://abenzerps/Spark-X2.5-4B-GGUF/Spark-X2.5-4B-Q4_K_M.gguf + sha256: 7934660bfc5b9bf04be0a0ac6179a1d16e1d4331b448857c86b8b2801b3ef72c +- !!merge <<: *spark-x2-5-4b + name: "spark-x2.5-4b-q6" + variants: [] + description: | + Spark-X2.5-4B in Q6_K GGUF format, with the embedded Jinja chat + template and a 32K-token default context. + overrides: + backend: llama-cpp + context_size: 32768 + known_usecases: + - chat + options: + - use_jinja:true + parameters: + model: llama-cpp/models/spark-x2.5-4b/Spark-X2.5-4B-Q6_K.gguf + temperature: 1 + top_p: 0.95 + top_k: -1 + min_p: 0 + repeat_penalty: 1 + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/spark-x2.5-4b/Spark-X2.5-4B-Q6_K.gguf + uri: huggingface://abenzerps/Spark-X2.5-4B-GGUF/Spark-X2.5-4B-Q6_K.gguf + sha256: 7293e99081e032b30e481c159aa35b3f7d133a746c1096910024ea71fff82246 +- !!merge <<: *spark-x2-5-4b + name: "spark-x2.5-4b-q8" + variants: [] + description: | + Spark-X2.5-4B in Q8_0 GGUF format, with the embedded Jinja chat + template and a 32K-token default context. + overrides: + backend: llama-cpp + context_size: 32768 + known_usecases: + - chat + options: + - use_jinja:true + parameters: + model: llama-cpp/models/spark-x2.5-4b/Spark-X2.5-4B-Q8_0.gguf + temperature: 1 + top_p: 0.95 + top_k: -1 + min_p: 0 + repeat_penalty: 1 + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/spark-x2.5-4b/Spark-X2.5-4B-Q8_0.gguf + uri: huggingface://abenzerps/Spark-X2.5-4B-GGUF/Spark-X2.5-4B-Q8_0.gguf + sha256: 58a4fc627cc2b2cbea02f81fb22960938e86bf3e62a2b3ae01c55a678481d46b - &qwen3-8-9b name: "qwen3.8-9b-q4" variants: From 1e5bf6ffb0b63575ae3539d7a9098280daf19794 Mon Sep 17 00:00:00 2001 From: Nathan Roll <96995554+Nathan-Roll1@users.noreply.github.com> Date: Sat, 12 Sep 2026 16:51:19 -0700 Subject: [PATCH 2/2] [gallery] feat: add Orukeet to the NeMo speech backend (#11998) * [gallery] feat: add Orukeet to the existing NeMo speech backend Assisted-by: Codex:gpt-6 Signed-off-by: Nathan Roll * docs(nemo): remove model-specific instructions Keep the backend guide focused on model families, as requested by mudler. Remove the gallery limitation instead of restoring an outdated claim. Assisted-by: Codex:GPT-6 Signed-off-by: Nathan Roll --------- Signed-off-by: Nathan Roll Co-authored-by: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com> --- docs/content/features/nemo-speech-cpp.md | 2 -- gallery/index.yaml | 29 ++++++++++++++++++++++++ 2 files changed, 29 insertions(+), 2 deletions(-) diff --git a/docs/content/features/nemo-speech-cpp.md b/docs/content/features/nemo-speech-cpp.md index 45de0ad7d..3efd44740 100644 --- a/docs/content/features/nemo-speech-cpp.md +++ b/docs/content/features/nemo-speech-cpp.md @@ -271,8 +271,6 @@ you want NeMo ASR on an AMD or Intel GPU, use **silently cut**: generation simply stops at the limit and the truncated translation is returned as if it were complete. Translate a sentence or a paragraph at a time rather than a whole document. -- **There are no gallery entries yet.** Models have to be converted with upstream's - converter and configured by hand, as below. This is a follow-up, not an oversight. ## Converting models diff --git a/gallery/index.yaml b/gallery/index.yaml index 068db623a..74610a5b1 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -49919,6 +49919,35 @@ - filename: parakeet-cpp/tdt-0.6b-v3-f16.gguf uri: huggingface://mudler/parakeet-cpp-gguf/tdt-0.6b-v3-f16.gguf sha256: 8ba47343e1e919895aca90e099150a01ed203ee0942d8ed31e27295efc5abb22 +- name: orukeet + url: github:mudler/LocalAI/gallery/virtual.yaml@master + urls: + - https://huggingface.co/oruk/orukeet + - https://github.com/Oruk-AI/orukeet + description: | + Orukeet is a 0.6B, 25-language fine-tune of Parakeet TDT v3 by Oruk. + Q8 GGUF for the nemo-speech-cpp backend. Runs locally on CPU, with + optional GPU acceleration through the backend's gpu option. + license: cc-by-sa-4.0 + tags: + - parakeet + - nemo-speech-cpp + - asr + - speech-recognition + - stt + - gguf + - multilingual + overrides: + backend: nemo-speech-cpp + known_usecases: + - transcript + name: orukeet + parameters: + model: orukeet/orukeet-v0.1.0-q8.gguf + files: + - filename: orukeet/orukeet-v0.1.0-q8.gguf + uri: https://huggingface.co/oruk/orukeet/resolve/555136b50265a132d4cea0d35560c26fc4f657ab/orukeet-v0.1.0-q8.gguf + sha256: 93ce19c6d8244acbfea980eeaf970531d4f216171578ef8e041dcc2d070a45bd - name: parakeet-cpp-ctc-1.1b url: github:mudler/LocalAI/gallery/virtual.yaml@master urls: