From 354bf5debb1fdf7f194b69fee49e435a495caebf Mon Sep 17 00:00:00 2001 From: "LocalAI [bot]" <139863280+localai-bot@users.noreply.github.com> Date: Sat, 18 Oct 2025 13:48:19 +0200 Subject: [PATCH] chore(model gallery): :robot: add new models via gallery agent (#6536) Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: mudler <2420543+mudler@users.noreply.github.com> --- gallery/index.yaml | 30 ++++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+) diff --git a/gallery/index.yaml b/gallery/index.yaml index 06f2626f1..7388eb7e0 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -22139,3 +22139,33 @@ - filename: Qwen-SEA-LION-v4-32B-IT.i1-Q4_K_M.gguf sha256: 66dd1e818186d5d85cadbabc8f6cb105545730caf4fe2592501bec93578a6ade uri: huggingface://mradermacher/Qwen-SEA-LION-v4-32B-IT-i1-GGUF/Qwen-SEA-LION-v4-32B-IT.i1-Q4_K_M.gguf +- !!merge <<: *qwen3 + name: "zirel-2-i1" + urls: + - https://huggingface.co/mradermacher/Zirel-2-i1-GGUF + description: | + **Model Name:** Zirel-2 + **Base Model:** Qwen/Qwen3-30B-A3B-Instruct-2507 (Mixture-of-Experts) + **Author:** Daemontatox + **License:** Apache 2.0 + + **Description:** + Zirel-2 is a highly capable, efficiency-optimized fine-tuned language model based on Qwen's 30B MoE architecture. It leverages only ~3.3B active parameters per inference step, delivering dense-model performance while minimizing resource usage. Designed for high reasoning, code generation, and long-context tasks (up to 262K tokens), it excels as a smart, responsive assistant. Ideal for deployment on consumer hardware or resource-constrained environments. + + **Key Features:** + - Mixture-of-Experts (MoE) design for efficiency + - 30.5B total parameters, 3.3B active per inference + - Long context (262,144 tokens) + - Optimized for reasoning, instruction-following, and creative generation + - Available in GGUF format for local inference + + **Use Case:** Personal AI assistant, code & content generation, complex reasoning tasks. + + *Note: The GGUF version in `mradermacher/Zirel-2-i1-GGUF` is a quantized derivative; the original model is `Daemontatox/Zirel-2`.* + overrides: + parameters: + model: Zirel-2.i1-Q4_K_S.gguf + files: + - filename: Zirel-2.i1-Q4_K_S.gguf + sha256: 9856e987f5f59c874a8fe26ffb2a2c5b7c60b85186131048536b3f1d91a235a6 + uri: huggingface://mradermacher/Zirel-2-i1-GGUF/Zirel-2.i1-Q4_K_S.gguf