From f9e9b2fe85e081bf80b04307b8a08de33a56b666 Mon Sep 17 00:00:00 2001 From: localai-org-maint-bot <306269227+localai-org-maint-bot@users.noreply.github.com> Date: Fri, 9 Oct 2026 20:05:54 +0000 Subject: [PATCH] chore(gallery): add Xing4.0 GGUF variants Offer Q5_K_M, Q6_K, and Q8_0 alongside the existing Q4_K_M build. Pin all four assets and correct the gallery description. Document direct installation of a quantization variant. Assisted-by: Codex:gpt-6 --- docs/content/features/model-gallery.md | 8 ++ gallery/index.yaml | 127 ++++++++++++++++++++++--- 2 files changed, 122 insertions(+), 13 deletions(-) diff --git a/docs/content/features/model-gallery.md b/docs/content/features/model-gallery.md index a5af4a904d2a..0d6312df6e96 100644 --- a/docs/content/features/model-gallery.md +++ b/docs/content/features/model-gallery.md @@ -547,6 +547,14 @@ ones, or both. Sizes are measured from the model's weights rather than downloaded, and cached. +For example, `xing4.0-29b-a4b` offers Q4_K_M, Q5_K_M, Q6_K, and Q8_0 +GGUF builds through llama.cpp. All four builds enable multi-token prediction. +To install a specific build directly, use its gallery name: + +```bash +local-ai models install xing4.0-29b-a4b-q6 +``` + The gallery listing only flags which entries offer variants, with a `has_variants` field. It deliberately does not describe them: measuring a variant is a network round trip per referenced build, so describing every diff --git a/gallery/index.yaml b/gallery/index.yaml index 23353a9bc9b5..2ae95e38e9da 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -69614,27 +69614,23 @@ sha256: ddf46c21d7078e95338cfc22306b19b276a29a5ad089023449dd54d4b6170a51 uri: https://huggingface.co/EldanRing/Winnow-E4B/resolve/b29f0a445f0b2b6207300b1bc6c3bed23ada8c91/gguf/mmproj-Winnow-E4B.gguf - name: xing4.0-29b-a4b + variants: + - model: xing4.0-29b-a4b-q5 + - model: xing4.0-29b-a4b-q6 + - model: xing4.0-29b-a4b-q8 url: github:mudler/LocalAI/gallery/virtual.yaml@master urls: + - https://huggingface.co/XingChen-AGI/Xing4.0-29B-A4B - https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF description: | - # Xing4.0-29B-A4B - - > [!Note] - > This repository provides the model weights and configuration files for Xing4.0-29B-A4B in Hugging Face Transformers format, compatible with Transformers, vLLM, SGLang, KTransformers, and other mainstream inference frameworks. - - **Xing4.0-29B-A4B** is a next-generation large language model in the Xing series (formerly TeleChat), developed by China Telecom Artificial Intelligence Technology Co., Ltd. With 29B total parameters and only 4B activated per token, it natively supports a 256K context length, extensible to 512K. It is the first model of this scale trained entirely on the Ascend NPU platform with the MindSpore framework, and deeply optimized for complex engineering tasks. - - For more information, please refer to our GitHub repository. - - ## Highlights - - ... + Xing4.0 is a 29B mixture-of-experts model with 4B active parameters for reasoning, coding, and tool use. + This Q4_K_M GGUF build uses llama.cpp with the embedded chat template and multi-token prediction. license: apache-2.0 tags: - llm - gguf - mtp + last_checked: "2026-10-09" overrides: backend: llama-cpp function: @@ -69655,7 +69651,112 @@ files: - filename: llama-cpp/models/Xing4.0-29B-A4B-Q4_K_M/Xing4.0-29B-A4B-Q4_K_M.gguf sha256: 09148c314610aa3f341b64023cf806342ee7702ba8857cfdfecfc09443de71f2 - uri: https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF/resolve/main/Xing4.0-29B-A4B-Q4_K_M.gguf + uri: https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF/resolve/845d645bfa2be61b7774f2a12a8bec9e2231ec5d/Xing4.0-29B-A4B-Q4_K_M.gguf +- name: xing4.0-29b-a4b-q5 + url: github:mudler/LocalAI/gallery/virtual.yaml@master + urls: + - https://huggingface.co/XingChen-AGI/Xing4.0-29B-A4B + - https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF + description: | + Xing4.0 is a 29B mixture-of-experts model with 4B active parameters for reasoning, coding, and tool use. + This Q5_K_M GGUF build uses llama.cpp with the embedded chat template and multi-token prediction. + license: apache-2.0 + tags: + - llm + - gguf + - mtp + last_checked: "2026-10-09" + overrides: + backend: llama-cpp + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + options: + - use_jinja:true + - spec_type:draft-mtp + - spec_n_max:6 + - spec_p_min:0.75 + parameters: + model: llama-cpp/models/Xing4.0-29B-A4B-Q5_K_M/Xing4.0-29B-A4B-Q5_K_M.gguf + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/Xing4.0-29B-A4B-Q5_K_M/Xing4.0-29B-A4B-Q5_K_M.gguf + sha256: 5dd24e9330791bb74de51488a96c9402fec5cb552115bc1a3ff29ddc28d83cde + uri: https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF/resolve/845d645bfa2be61b7774f2a12a8bec9e2231ec5d/Xing4.0-29B-A4B-Q5_K_M.gguf +- name: xing4.0-29b-a4b-q6 + url: github:mudler/LocalAI/gallery/virtual.yaml@master + urls: + - https://huggingface.co/XingChen-AGI/Xing4.0-29B-A4B + - https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF + description: | + Xing4.0 is a 29B mixture-of-experts model with 4B active parameters for reasoning, coding, and tool use. + This Q6_K GGUF build uses llama.cpp with the embedded chat template and multi-token prediction. + license: apache-2.0 + tags: + - llm + - gguf + - mtp + last_checked: "2026-10-09" + overrides: + backend: llama-cpp + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + options: + - use_jinja:true + - spec_type:draft-mtp + - spec_n_max:6 + - spec_p_min:0.75 + parameters: + model: llama-cpp/models/Xing4.0-29B-A4B-Q6_K/Xing4.0-29B-A4B-Q6_K.gguf + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/Xing4.0-29B-A4B-Q6_K/Xing4.0-29B-A4B-Q6_K.gguf + sha256: ebbb73a92993c2de6aa1a089515c287b5ba7d58e40f91b0bbaca55e81054c71b + uri: https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF/resolve/845d645bfa2be61b7774f2a12a8bec9e2231ec5d/Xing4.0-29B-A4B-Q6_K.gguf +- name: xing4.0-29b-a4b-q8 + url: github:mudler/LocalAI/gallery/virtual.yaml@master + urls: + - https://huggingface.co/XingChen-AGI/Xing4.0-29B-A4B + - https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF + description: | + Xing4.0 is a 29B mixture-of-experts model with 4B active parameters for reasoning, coding, and tool use. + This Q8_0 GGUF build uses llama.cpp with the embedded chat template and multi-token prediction. + license: apache-2.0 + tags: + - llm + - gguf + - mtp + last_checked: "2026-10-09" + overrides: + backend: llama-cpp + function: + automatic_tool_parsing_fallback: true + grammar: + disable: true + known_usecases: + - chat + options: + - use_jinja:true + - spec_type:draft-mtp + - spec_n_max:6 + - spec_p_min:0.75 + parameters: + model: llama-cpp/models/Xing4.0-29B-A4B-Q8_0/Xing4.0-29B-A4B-Q8_0.gguf + template: + use_tokenizer_template: true + files: + - filename: llama-cpp/models/Xing4.0-29B-A4B-Q8_0/Xing4.0-29B-A4B-Q8_0.gguf + sha256: 2ebec362780b96c299f5a062966dcab41de41e96fbab92eee3f4e2b35f623244 + uri: https://huggingface.co/Venastine-Research/Xing4.0-29B-A4B-GGUF/resolve/845d645bfa2be61b7774f2a12a8bec9e2231ec5d/Xing4.0-29B-A4B-Q8_0.gguf - name: artemis-31b-v1.2-q4 variants: - model: artemis-31b-v1.2-q8