##新闻 官方技能来改善快速写作:GitHub上的技能

<div align="center"> ![模型请求](https://privacy。

”””引用这份工作。

<div> <p style="margin: 0 0 0px 0; margin-top: 0px;"> <em><a href="https://unsloth.

<!-- markdownlint-disable first-line-h1 --> <!-- markdownlint-disable html --> <!-- markdownlint-disable no-duplicate-header --> <div align="center"> <img src="https://github.

点击此处查看 4B 版本. 也查看 DeepThought,我的科学 finetune 基于此模型: xtools-at/Qwen3.5-9B-DeepThought-STEM

异端 v1.2.0+ 使用任意排位(ARA)方法定制

Anima-Base - 预先训练的,未精确的基础模型。

## 性能 计量 此模型 原始模型(Qwen/Qwen3)

这是NVIDIA/QWEN3的组合。

- **公告: **Cloudflare博客上的关键决策模型 - **决策指数领导板: **关键评级。

这是一个未经审查的Qwen/Qwen3版本。

- 介绍 - 模型一眼 - 突出 - 建筑 - 基准 - 快速启动 - 推断 - 服务和生态系统 - 推荐参数 - 评估细节 - 引用 -...

我们正在发布 **MiniCPM5-2B**,这是 **MiniCPM5**系列中的第二个模型,继 MiniCPM5-1B之后。

**VoxCPM2**是一个无代币的、散发性自转式文本对话模型 - **2B参数**, **30种语言**, **48kHz**音频输出,训练超过 **2百万小时**的多语言音频。

<div> <p style="margin: 0 0 0px 0; margin-top: 0px;"> <em><a href="https://unsloth.

人造全身运动跟踪与一个可区分的世界模型。

LFM3-1.

> !NOTE] > [2026.02.06] 我们开源的实时Web演示可部署在您自己的设备上,如Mac或GPU。

低于1的KLD是优良的,零是完美的。

」小小二。

* 2026.05.17] ⭐️⭐️⭐️ 我们发布了 MiniCPM-V 4.6 的 API 服务,同时使用一个 **公共免费的 API 密钥**! 试试。

** 使用 LightX2V Studio 在线创建 - 从文本、图像或参考材料中生成视频。

我们正在发布 **MiniCPM5-1B**,这是 **MiniCPM5**系列中的第一个模型。

**MiniCPM-V 4号。

- **公告: **Cloudflare博客上的关键决策模型 - **决策指数领导板: **关键评级。

“Nemotron 3 Diarization”是一个开放重量级扬声器 diarization 模型,旨在确定“谁在什么时候说话”在现实世界的音频。

VoxCPM 是一款无代币的 Text-to-Speech (TTS) 系统,在语音合成中重新定义现实主义。

”““你要精确、准确。

##在线API 直接通过 API 使用 MiniMax-H3-Image。

用于MiniMax H3的实验 LoRA,为ComfyUI包装,主要用“ref2va”重量测试。

SmolVLM是一款紧凑的开放型多式模型,可接受图像和文本输入的任意序列,以产生文本输出。

一个经过培训的模型,在Qwen3上使用了几轮SFT和RLOO。

主模型卡: google/gemma-4-E2B-it 此模型卡以 Android、iOS、桌面、物联网和 Web 上部署的方式提供 Gemma 4 E2B 模型卡。

d1-3B 是基于 LFM2 的 3B 参数 ** 决策模型。

测量在 **2026-09-24**与H200 BF16推断和严格的编码。

### Z-Image Base上的完整动漫精彩调节 - 丰富的美学,强大的多样性,完整的负面提示支持 <a href="https://huggingface。

** 增强的现实主义:** 更自然的进展和表达式的交付,基于预览。

2026-11-10 — 2D v1 发布 > > 添加了 **`minimax_h3_latent_upscaler_2d_conv_v1`**: > - **一半的参数** — 182,140,312 参数, **52.

lfm2的。

<div align="center"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://framerusercontent.

**MiniCPM-V 4号。

提供两组文件。

• **SOTA文档解析**: **93。

<img width="400px" src="https://qianwen-res.

SmolVLM-256M是世界上最小的多式车型。

◎Cosmos3 ‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧... NVIDIA CosmosTM 是一个旨在加速物理人工智能的发展,通过...

<div> <p style="margin-top: 0;margin-bottom: 0;"> <em>See <a href="https://unsloth.

我们正在发布 **MiniCPM5-2B**,这是 **MiniCPM5**系列中的第二个模型,继 MiniCPM5-1B之后。

此分类上一篇: (https)

LFM2是一种用于设备部署的混合模型家族。

此存储库包含模型重量和训练后的模型配置文件,以Hugging Face Transformers格式。

Kromoss 是 MOSS-Image-Base 的 1024 像素的完整调节。

马云2号

介绍 - 模型信息 - 评估结果 - 快速启动 - 伴侣数据集 - 许可证和引用

<img src="https://cdn-uploads.

此目录包含音频。

<p align="center"> <img src=".

### 超快的8步生成 - 光现实主义和动画检查点在一个存储库中,一对一 <a href="https://huggingface。

◎Cosmos3 ‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧... NVIDIA CosmosTM 是一个旨在加速物理人工智能的发展,通过...

## - 帮助菜单和互动会话

![Try the Playground](https://chatrlabs.

**SenseNova-Loop-9B-Preview是为长期背景和代理工作流程开发的多式语言模型。

通过下载或使用这些权重,您接受MiniMax H3社区许可协议。

d1-2B 是文本、图像和视频的决策模型,构建在 LFM3-VL-2B 上。

Edge0 不是重新标记的 Qwen3。

LFM3-VL-2B 是用于设备上的视觉理解的 **视觉语言模型。

• 运行时间: SGLang 0

关键字:Qwen / Qwen3

我们正在发布 **MiniCPM5-1B**,这是 **MiniCPM5**系列中的第一个模型。

<div align="center" style="line-height: 1;"> <a href="https://huggingface.

### 模型描述 <!-- 提供一个更长的概述这个模型是什么。 --> - ** 由:** [HeartMuLa Team] - ** 许可证:** [Apache 2.0]

Swift Flash Next 的混合精度 GGUF 量化,具有 Swift 特定的 GSQ 精炼和重复使用的 ISTA GSQ-RCO per-tensor 配置配置配置配置。

<div align="center" style="line-height: 1;"> <a href="https://developer.

主模型卡: google/gemma-4-E4B-it 此模型卡以 Android、iOS、桌面、物联网和 Web 上部署的方式提供 Gemma 4 E4B 模型卡。

低于1的KLD是优良的,零是完美的。

◎Cosmos3 ‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧... NVIDIA CosmosTM 是一个旨在加速物理人工智能的发展,通过...

<div align="center"> <img class="block dark:hidden" src="https://raw.

异端 v1.2.0+ 使用任意排位(ARA)方法定制

该模型已转换为 MLX 格式从 `Qwen/Qwen3.8-27B` 使用 mlx-vlm 版本 **0.6.8**. 请参阅原始模型卡以获取模型的详细信息。

1. 介绍 2. 架构 3. 基准 - 知识和编码 - 推理和数学 - 代理 4. 引用 - 拥抱面部 - vLLM - SGLang 5. 脚注 6. 引用

![早期获取LotusMake订阅者](https://LotusMake。

此分类上一篇: <div align="center"> <img src="https://raw.

一八。

重新包装的模型文件 ComfyUI. 原始模型存储: https://huggingface.co/MiniMaxAI/MiniMax-H3-Image

这是一个未经审查的Qwen/Qwen3版本。

”““你要精确、准确。

校准跟踪: md JSON 安全传感器

SmolVLM2-500M-Video是一个轻量级的多式模型,旨在分析视频内容。

SmolVLM-500M是一款微小的多式车型,是SmolVLM家族的一员。

在 Llama 应用程序中运行: https://llama.app 或从终端: ``bash llama serve -hf ggml-org/gpt-oss-20b-GGUF ``

<!-- ### quantize_version: 2 --> <!-- ### output_tensor_quantised: 1 --> <!-- ### convert_type: hf --> <!-- ### vocab_type: --> <!-- ### tags: nicoboss --> <!-- ### quants: Q2_K IQ3_M Q4_K_S...

◎Cosmos3 ‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧... NVIDIA CosmosTM 是一个旨在加速物理人工智能的发展,通过...

<p align="center"> <img src="https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/ZCLLXZk2LQBG0YH_BmiIN.gif" style="width:100%; max-width:1000px; height:auto;"> </p>

!Qwen3-4B Z-Image-Engineer V2: The Z-Engineer Returns,为LM Studio,ComfyUI和MLX,由BennyDaBall **跟随我X @BennyDaBall_OG !** **想支持我的工作?在Ko-fi上给我买咖啡!**

标签: Turbo Brilliance 系统 V1.0

> !NOTE] > [2026.02.06] 我们开源的实时Web演示可部署在您自己的设备上,如Mac或GPU。

# 模型摘要

# <a href="https://huggingface.

![HuggingFace](https://huggingface.

TildeOpen LLM是一个开源的基础(基础)语言模型,旨在服务不足的北欧和东欧语言。

每体重 2.05 位 / H4, NG4 每体重 3.05 位 / H5, NG5 每体重 4.05 位 / H6, NG6 每体重 5.05 位 / H6, NG6 每体重 6.05 位 / H6, NG6

**Licon VBVR** 是 **MiniMax-H3** 的图像对视频 LoRA,是针对 H3 进行重建和重新处理的 VBVR 数据的培训。

!Z-Engineer V4 SMART 培训, 4B 全细调, Z-Image Turbo 编码器,由BennyDaBall **跟随我X @BennyDaBall_OG !** **想支持我的工作?在Ko-fi上给我买咖啡!**更新:

与 llama 进行本地运行。

◎Cosmos3 ‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧... NVIDIA CosmosTM 是一个旨在加速物理人工智能的发展,通过...

** 将文档页面转化为结构化信息。

所有模型均对 **BF16**基模型进行评估。

### girlslikekrea2_gzl !

在 Llama 应用程序中运行: https://llama.app 或从终端: ``bash llama serve -hf ggml-org/Qwen3.8-27B-GGUF ``

## ∴ V2 涡轮光亮度系统。

- **公告:**Cloudflare博客上的Clef-omni模型重现Clef-Omni是一个30B-A3B专家组合的多式模型,将一个状态和一个编写问题的方案转化为决策。

!AEON Qwen — Supreme Being of the Digital Cosmos ![Buy me a coffee on LotusMake](https://lotusmake.

重新包装为 ComfyUI 的模型文件。

”““你要精确、准确。

低于1的KLD是优良的,零是完美的。

### 180B 专家混合 · 视觉语言 · **#1 在十个Hugging Face官方领导板上**:九个由这个模型(MDPBench 83)。

长的MiniMax-H3视频从单个提示器同步音频,在ComfyUI。

重新包装为 ComfyUI 的模型文件。

!Qwen3-4B Z-Image Turbo Abliterated V1, drop-in text encoder, GGUF, by BennyDaBall **Follow me on X @BennyDaBall_OG !** **Want to support my work? Buy me a coffee on Ko-fi !** **"I'm sorry,...

每个层面都需要视觉投影机 - `--model` 接收了 LLM GGUF, `--mmproj` 接收了投影机: Fh 文件, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸, 尺寸。

Blog | GitHub This repository contains the DFlash 2 draft model for `zai-org/GLM-5.

- About This Build - Introduction - Model at a Glance - Highlights - Architecture - Benchmarks - Quickstart - Inference - Serving & Ecosystem - Recommended Parameters - Evaluation Details -...

一个“Qwen3”的九个GGUF文件的家族。

此模型卡是 neroued/Qwen3 的版本控制源。

This repository contains the MiniCPM-o 2.

SmolVLM2-256M-Video is a lightweight multimodal model designed to analyze video content.

<p align="center"> <img src="https://cdn-uploads.

JoyAI-Video-Edit 是一个实时、指导视频编辑系统,用于开放式视频流。

<details> <summary> 点击 <b>here</b> 查看思维过程的文本墙。

提供两组文件。

<p align="center"> <a href="https://huggingface.

在 Llama 应用程序中运行: https://llama.app 或从终端: ``bash llama serve -hf ggml-org/gemma-4-E4B-it-GGUF ``

d1-Omni-2B is a **decision model for text, pictures, and audio**, built on LFM3-Omni-2B.

马云:九九。

<!-- orukeet-team:start --> <p> Nathan Roll^1,2 · Irene Yi^1,2 · Büşra Marşan^1,2<br> Vianney Grenez¹ · Gabriel Stein⁴ · Momcilo...

<p align="center"> <img src="assets/logo.

<p style="line-height: 100%;边缘底部: 0cm;字体重:正常"> <br/> </p> <p><span style="display: inline-block;边界: none; 贴纸: 0cm,字体重:正常"> <br/> </p><p><span style="display: inline-block; border: none; 贴纸: 0cm,字体重: normal="#000000,字体大小="3"...

**A tiny embedding model that punches far above its weight.

在 Llama 应用程序中运行: https://llama.app 或从终端: ``bash llama serve -hf ggml-org/gemma-4-12B-it-GGUF ``

**KaLM-Embedding-Gemma3-12B-2511** is a versatile and compact embedding model, which achieves SOTA performance in MMTEB (due to 11-2025).

<p align="center"> <a href="https://huggingface.

## 性能 计量 此模型 原始模型(Qwen/Qwen3)

培训代码 - 评估代码 - 所有培训数据 K2-Horizon-0.

TiRex is a 35M parameter pre-trained time series forecasting model based on xLSTM.

在 Llama 应用程序中运行: https://llama.app 或从终端: ``bash llama serve -hf ggml-org/gemma-4-26B-A4B-it-GGUF ``

### Main Benchmarks | Benchmark | Score | | --- | --- | | HellaSwag (0-shot) | 83.

These models were evaluated against a large collection of different datasets and metrics to cover different aspects of text generation.

在 Llama 应用程序中运行: https://llama.app 或从终端: ``bash llama serve -hf ggml-org/gemma-4-E2B-it-GGUF ``

Here is a code snippet with `apply_chat_template` to show you how to load the tokenizer and model and how to generate content.

异端 v1.2.0+ 使用任意排位(ARA)方法定制

![Try the Playground](https://chatrlabs.

It's now late **January**, **2026**, open source is crushing closed frontier (Kimi K2.

<details> <summary>Click <b>here</b> for a wall of text.

Korean-tuned 35B local AI for PCs: choose a full-model GGUF build and run without a GPU.

!AEON Qwen — Supreme Being of the Digital Cosmos ![Buy me a coffee on LotusMake](https://lotusmake.

This finetune was an absolute pain to make.

- **TaoMate H3 FL2VA 3-Step:**在WanGP中,已过时的FL2VA加速器配置器使用了新的Rank-128适配器。

该模型已转换为 MLX 格式从 `Qwen/Qwen3.8-27B` 使用 mlx-vlm 版本 **0.6.8**. 请参阅原始模型卡以获取模型的详细信息。

## More than 10000 members strong 💪 A hub for users and makers alike! ---

MNN-Pocket is a family of compact language models co-designed with the MNN inference stack for **on-device deployment**.

!Z-Image-Engineer V6 SMART DoRA, prompt writer + Z-Image text encoder, by BennyDaBall **Follow me on X @BennyDaBall_OG !** **Want to support my work? Buy me a coffee on Ko-fi !**

此模型卡是 neroued/Qwen3 的版本控制源。

在 Llama 应用程序中运行: https://llama.app 或从终端: ``bash llama serve -hf ggml-org/gpt-oss-120b-GGUF ``

这是一个未经审查的Qwen/Qwen3版本。

**Beta7 fix1 HQ (18.

Independently evaluated via the UGI benchmark, **Assistant_Pepe_70B** was **ranked 1st in the world**, combining exceptional intelligence and instruction-following capabilities with next to no...

This is a GGUF quantized version of Qwen-Image-2.

《画廊》

* 2026.05.17] ⭐️⭐️⭐️ 我们发布了 MiniCPM-V 4.6 的 API 服务,同时使用一个 **公共免费的 API 密钥**! 试试。

**Release status: Ready for release.

**MiniCPM-V 4号。

<p style="margin-bottom: 0cm"><br/> </p> <p style="border: none; padding: 0cm"><span style="display: inline-block; border: none; padding: 0cm">The sky scraper sways.

KLM 1.6 is a 99M-parameter base language model that continues training from a checkpoint of our KLM 1.5 taken before its decay stage.

PAN 2 is a general-purpose world model: it simulates how the world will change in response to actions.

!AEON Qwen — Supreme Being of the Digital Cosmos ![Buy me a coffee on LotusMake](https://lotusmake.

<img width="400px" src="https://qianwen-res.

![MiniMax-H3 video comparison preview — click to open the interactive demo](https://cdn.

Darwin-31B-Opus is a reasoning-enhanced model created by merging google/gemma-4-31B-it (Father) and TeichAI/gemma-4-31B-it-Claude-Opus-Distill (Mother) using the Darwin V6 engine.

Text-only RSI reasoning in a 111.

Main Model Card: google/gemma-4-12B-it This model card provides the Gemma 4 12B model in LiteRT-LM format ready for deployment on macOS, Linux and Windows, as well as web in a more limited capacity.

<p style="line-height: 100%; margin-bottom: 0cm"><br/> </p> <p><span style="display: inline-block; border: none; padding: 0cm"><i>Come up with six plots for a new &quot;Black Mirror&quot; episode...

</details> ---

<p align="center"> <a href="https://huggingface.

Edge0 不是重新标记的 Qwen3。

!Z-Image-Engineer V6 GGUF, for ComfyUI and LM Studio, by BennyDaBall **Follow me on X @BennyDaBall_OG !** **Want to support my work? Buy me a coffee on Ko-fi !** GGUF quantized release for...

This repository contains the **BF16 version of S5-nano** (9.

<img width="400px" src="https://qianwen-res.

The best open-sourced music generation model in terms of lyrics controllability and music quality.

### girlslikezi_fwf !

第一

These models were evaluated against a large collection of different datasets and metrics to cover different aspects of text generation.

Click here for the 9B version. Also check out DeepThought, my science finetune based on this model: xtools-at/Qwen3.5-4B-DeepThought-STEM

GGUF quantizations of deepseek-ai/DeepSeek-V4.

工程师V2

标签: Turbo Brilliance 系统 V1.0

Darwin-4B-David is the first second-generation (Generation 2) model in Darwin history — **a model evolved from an already-evolved model.

![Discord](https://discord.

GameScaling models are Qwen3.

!AEON Qwen — Supreme Being of the Digital Cosmos ![Buy me a coffee on LotusMake](https://lotusmake.

!AEON Ornith — Supreme Being of the Digital Cosmos ![Buy me a coffee on LotusMake](https://lotusmake.

A 三。

### Description Déjà View Looping Transformer (DVLT) is a feed-forward three-dimensional (3D) reconstruction model that takes unposed Red, Green, Blue (RGB) images or video and predicts per-pixel...

<p align="center"> <a href="https://huggingface.

Luxical-One is a small lexical-dense text embedding model distilled from `Snowflake/snowflake-arctic-embed-m-v2.

A small text-to-speech model for 11 Indian languages, with two voices: **Satvik** (male) and **Charvi** (female).

## Model Overview - **Model Architecture:** Qwen3_5ForConditionalGeneration - **Input:** Text / Image - **Output:** Text - **Model Optimizations:** - **Weight quantization:** INT4 -...

该模型提供了几种Qwen/Qwen2的变体。

Website: basalt.

THX-02 is a thinking decision model.

![Discord](https://discord.

我们正在发布 **MiniCPM5-2B**,这是 **MiniCPM5**系列中的第二个模型,继 MiniCPM5-1B之后。

`MiniCPM5-2B-DSpark` is a DSpark draft checkpoint trained for exact pairing with MiniCPM5-2B and its tokenizer.

A growing collection of **free Krea 2 character LoRAs**, with visual previews, trigger words, recommended strengths, and direct downloads.

该模型提供几种变体的deepseek-ai/DeepSeek-R1-Distill-Qwen-1.

!Qwenx86: Qwen writes C, the x86 computer in the weights compiles and executes it (actual recorded execution) ▶️ **See it in action: Qwenx86 Debugger Space**.

**Darwin-28B-REASON** is a reasoning-enhanced **standalone model** derived from **Darwin-28B-Opus**.

This is the int4 quantized version of **MiniCPM-o 2.6**. Running with int4 version would use lower GPU memory (about 9GB).

**MiniCPM-V 4号。

<!-- ### quantize_version: 2 --> <!-- ### output_tensor_quantised: 1 --> <!-- ### convert_type: hf --> <!-- ### vocab_type: --> <!-- ### tags: --> <!-- ### quants: x-f16 Q4_K_S Q2_K Q8_0 Q6_K...

![AEON Uncensored NVFP4 Mixed Precision Release](https://lotusmake.

The day that Anthropic released their "text watermark" embedded in their models, I give you: **ZeroZero Heretic** ZERO slop.

### Foresight before action.

## Strategic Blueprint | Exhaustive Version 1.0 ---

重新包装为 ComfyUI 的模型文件。

**Krea 2 AnyPaint provides arbitrary-mask inpainting, outpainting, and image editing for Krea 2 Turbo.

TiRex-2 is a **pretrained time series foundation model** that forecasts one or many target variates directly from their history, optionally conditioned on past and future-known covariates.

> 🎵 **Try HeartMuLa online — no GPU or setup needed:** heartmula.

**Paper** | **Code** | **Project Page**

Darwin-9B-Opus is a 9B dense parameter reasoning model created using Darwin V5.

## Model Overview - **Model Architecture:** Glm5NextForConditionalGeneration - **Input:** Text / Image - **Output:** Text - **Model Optimizations:** - **Weight quantization:** FP4 -...

> **New:** For arbitrary masks, unified inpainting/outpainting, disconnected regions, > and boundary-crossing edits, see **Krea 2 AnyPaint**.

<p align="center"> <a href="https://huggingface.

在 Llama 应用程序中运行: https://llama.app 或从终端: ``bash llama serve -hf ggml-org/gemma-4-31B-it-GGUF ``

| Property | Value | |---|---| | Architecture | Qwen3.

## 大家好,我是Furkan Gözükara博士,我是一名计算机工程师,担任助理教授 + 全日制人工智能研究员 + 开发人员 + 教程制作人

我们正在发布 **MiniCPM5-1B**,这是 **MiniCPM5**系列中的第一个模型。

![GitHub](https://github.

meet the bees |tiny bees |mood bees |bee hives |delve bees |get started |model details

This repository contains the Multi-Token Prediction (MTP) drafter weights split from `Qwen/Qwen3.

--- ElasticModels是由TheStage AI ANNA生产的模型:自动神经网络加速器。

## 模型概述 - ** 模型架构:** Qwen4ExpForConditionalGeneration - ** 输入:** 文本 / 图像 / 视频 - ** 输出:** 文本 - ** 模型优化:** - ** 重量量化:** FP4 -...

超级V4

RFD4-Proteína is a flow-matching model for biomolecular design.

**67.

### Judgment at the point of action.

Demo | Release Announcement | Technical Overview | Technical Report Mobile-oriented models for decoding swipe gestures into text.

The license of the quantized files follows the license of the original model: - Qwen RESEARCH LICENSE AGREEMENT: https://huggingface.

> **Important:** The Hugging Face auto-generated Diffusers snippet and ordinary > style-LoRA workflows are not compatible with this functional adapter.

### 180B 专家混合 · 视觉语言 · **#1 on ExtractBench (90)。

<p align="center"> <img src="assets/logo.

<div align="center"> ![Try the live demo — open in browser](https://huggingface.

Run it in the Llama app: https://llama.

OrcaRouter's Qwen3.

**Fast Thai + English printed-text OCR that runs on a CPU.

主模型卡: google/embeddinggemma-2 EmbeddingGemma 2 740M是 EmbeddingGemma 2 家族的完整,本地多式变体,用于低延迟,使用 LiteRT 进行设备推断。

**Qwen3.

Run it in the Llama app: https://llama.

Compact weights and a native CIRU runtime for AMD Strix Halo.

# Model Overview

Crowd-v1 是社区驱动的分布式 AI 项目 **CrowdGPT**的第一个正式模型架构。

!banner ```python pip install inno-kokoro ``` Zero-shot voice tuner for Kokoro-82M. Outputs base Kokoro compatible voice packs @ `510, 1, 256]`. ([write-up)

◎Cosmos3 ‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧... NVIDIA CosmosTM 是一个旨在加速物理人工智能的发展,通过...

**Towards Compact and Efficient Video Autoencoders** Wan2.

Developed by iApp, launched with AIEAT.

Thai document pages to Markdown, with tables as HTML.

!AEON Qwen — Supreme Being of the Digital Cosmos ![Buy me a coffee on LotusMake](https://lotusmake.

Run it in the Llama app: https://llama.app Or from the terminal: ```bash llama serve -hf ggml-org/Qwen3.8-Flash-Next-GGUF ```

<div align="center"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://framerusercontent.

This repository contains several checkpoints derived from the **MiniMax H3 3D Latent Upscaler** by LBH-123-AI.

我们正在发布 **MiniCPM5-1B**,这是 **MiniCPM5**系列中的第一个模型。

Explained here ---

### *Character fidelity engineered to survive deep-context drift.

!Krea-2-Engineer V1 GGUF Prompt Writer,用于LM Studio和Lama。

**Darwin-Prediction-Weather** is VIDRAFT's weather prediction system.

Run it in the Llama app: https://llama.app Or from the terminal: ```bash llama serve -hf ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF ```

第一

◎Cosmos3 ‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧... NVIDIA CosmosTM 是一个旨在加速物理人工智能的发展,通过...

**Darwin V8 시리즈의 2B 경량 모델** Claude Opus 4.5/4.6 및 Sonnet 4.6의 추론 스타일을 주입한 Qwen3.5-2B 기반 모델. ---

<p align="center"> <a href="https://huggingface.

**Dials** are ~14 KB steering files made from prompt pairs, with **no training**.

A **13.

> 🎵 **Try HeartMuLa online — no GPU or setup needed:** heartmula.com > Make songs from lyrics, create covers and remixes, and edit any passage of a song in MuLa Studio. Free daily credits.

Darwin-4B-Opus is a reasoning-enhanced model created by merging google/gemma-4-E4B-it (Father) and arsovskidev/Gemma-4-E4B-Claude-4.

LiquidCodec is a **compact, streaming audio codec** that compresses speech into discrete tokens and converts them back into audio.

If TeraTTSv2 is useful to you, consider supporting its development.

![Buy me a coffee on LotusMake](https://lotusmake.

Project page · Paper · Evaluation code · PhyGround dataset PhyJudge-9B is a PEFT/LoRA adapter for Qwen3.

* 2026.05.17] ⭐️⭐️⭐️ 我们发布了 MiniCPM-V 4.6 的 API 服务,同时使用一个 **公共免费的 API 密钥**! 试试。

Edge0 不是重新标记的 Qwen3。

GGUF version of MiniCPM5-2B-DSpark for speculative decoding with MiniCPM5-2B in llama.cpp. | File | Precision | Size | | --- | --- | --- | | `MiniCPM5-2.6B-DSpark.gguf` | BF16 | 653 MB |

我们正在发布 **MiniCPM5-2B**,这是 **MiniCPM5**系列中的第二个模型,继 MiniCPM5-1B之后。

The license of the quantized files follows the license of the original model: - MiniMax-H3: https://huggingface.

A video System One model based on Qwen/Qwen3.5-4B, using direct candidate scoring for **temporal grounding** and **multiple-choice video QA**. Paper · Homepage · Code · Usage guide

<p align="center"> <a href="https://arxiv.

Hey there! 👋 Meet **Forge-1-E2B**.

- **2026-10]** Inference code released. - **[2026-09]** Paper released on [arXiv.

- **CyberBase** is a lmsys/vicuna-13b-v1.5-16k QLORA fine-tuned on CyberNative/github_cybersecurity_READMEs with a single 3090. - It might, therefore, have inherited prompt template of FastChat

Run it in the Llama app: https://llama.app Or from the terminal: ```bash llama serve -hf ggml-org/Qwen3.6-35B-A3B-GGUF ```

Run it in the Llama app: https://llama.

**Identity Lock:** Cross-industry strategic ideation engine.

我们正在发布 **MiniCPM5-2B**,这是 **MiniCPM5**系列中的第二个模型,继 MiniCPM5-1B之后。

**A decision model for routing AI agents across cloud and local models.

!AEON Ornith — Supreme Being of the Digital Cosmos ![Buy me a coffee on LotusMake](https://lotusmake.

> !NOTE] > **Updated 2026-11-11:** Re-quantized with oMLX v0.

<details> <summary>Wolf Defender family integrations</summary> <table> <tr> <td align="center"><a href="https://camel.

Run it in the Llama app: https://llama.app Or from the terminal: ```bash llama serve -hf ggml-org/Qwen3.6-27B-GGUF ```

!Amos 1 large: 94.

GGUF conversions of tencent/Qwen3.

> 🎵 **Try HeartMuLa online — no GPU or setup needed:** heartmula.

Run it in the Llama app: https://llama.app Or from the terminal: ```bash llama serve -hf ggml-org/Qwen3.5-0.8B-GGUF ```

该模型已转换为 MLX 格式从 `Qwen/Qwen3.8-27B` 使用 mlx-vlm 版本 **0.6.8**. 请参阅原始模型卡以获取模型的详细信息。

# - source_model: g:/3B/Llama-3.

Prebuilt runtime binaries for Needle 3, Whistle and Mirage. One engine runs text, speech and vision locally. Model weights live in their own repositories.

GGUF conversion of **xtools-at/DeepSeek-V4-Pro-Qwen3.

**Multilingual PII Detection Model for European Languages** A state-of-the-art multilingual model for detecting Personally Identifiable Information (PII) across 26 European languages (all EU...

This repository contains LiteRT-LM variants of Qwen3-0.

<div align="center" style="line-height: 1;"> <a href="https://developer.

| Resource | Link | |---|---| | **Model Weights + Full Documentation** | AEON-7/Gemma-4-26B-A4B-it-Uncensored-NVFP4 on HuggingFace | | **DFlash vLLM Container (DGX Spark)** | `ghcr.

Run it in the Llama app: https://llama.

Fight Flow V1 does not mainly teach new moves.

来自碳家族的DNA注释模型。

Blog | GitHub This repository contains the DFlash 2 draft model for `zai-org/GLM-5.

!LFM2.

## News 🔥 - `2026/11/06`: 🤗 Released the Kanana-2-v-decide-3b-2611 model weights.

LFM3-1.

Run it in the Llama app: https://llama.app Or from the terminal: ```bash llama serve -hf ggml-org/DeepSeek-V4-Flash-0731-GGUF ```

- **Unified vision-language model**: text, image, and video input on the Qwen3.

This model was obtained by using the following branch with LLM Compressor: https://github.com/vllm-project/llm-compressor/pull/2647

> !NOTE] > **SnapRef v2a stable.

### 模型描述 <!-- 提供一个更长的概述这个模型是什么。 --> - ** 由:** [HeartMuLa Team] - ** 许可证:** [Apache 2.0]

Quantized **GGUF** weights for **MiniMax-H3-Image**, converted directly from the structurally pruned BF16 master checkpoint provided by **Comfy-Org** (`minimax_h3_image_pruned_bf16.

This repository contains a **LoRA adapter** fine-tuned on the state-of-the-art **Qwen 3.

> !NOTE] > [2026.02.06] 我们开源的实时Web演示可部署在您自己的设备上,如Mac或GPU。

### Personal Anime Model built on Anima **Two Generations • Clean Anime Aesthetics • Tag + Natural Language • Anima-Compatible** **Diffusion Models • 1 MP Native • LoRA-friendly** <br> <a...

### Local, reversible PII detection for Italian legal text **_Use frontier models without giving up your data.

**A 4B Korean-reasoning model built with VIDRAFT's Chimera technology.** Most models get better by getting bigger. This one got better by **combining** — and then refining itself. ---

**🌐 Project website: nevo-project.

> **⚠️ Experimental.

## ∴ V2 涡轮光亮度系统。

Community packages for two separate jobs: Molmo2-8B INT8 ConvRot as an experimental MiniMax H3 encoder, and MolmoPoint-8B point tracking for video segmentation.

◎Cosmos3 ‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧‧... NVIDIA CosmosTM 是一个旨在加速物理人工智能的发展,通过...

我们正在发布 **MiniCPM5-2B**,这是 **MiniCPM5**系列中的第二个模型,继 MiniCPM5-1B之后。

我们正在发布 **MiniCPM5-1B**,这是 **MiniCPM5**系列中的第一个模型。

* 2026.05.17] ⭐️⭐️⭐️ 我们发布了 MiniCPM-V 4.6 的 API 服务,同时使用一个 **公共免费的 API 密钥**! 试试。

**Mixed-Precision FP4** quantization of Darwin-9B-Opus, built on NVIDIA Blackwell-native NVFP4.

!Qwen Image 2.

我们正在发布 **MiniCPM5-2B**,这是 **MiniCPM5**系列中的第二个模型,继 MiniCPM5-1B之后。

**Cosmos3** | **Cosmos3 Examples** | **Code** | **Technical Report** | **Website**

It is **HIGHLY RECOMMENDED** to use the **Roleplay \ Adventure format the model was trained on**, see the examples below for syntax.

> **비드래프트(VIDRAFT) 한국 법률 특화 28B 멀티모달 언어 모델** > Darwin family 한국어 도메인 변종 — 법률 분야 ---

A 4-bit build of Qwen3.

Run it in the Llama app: https://llama.app Or from the terminal: ```bash llama serve -hf ggml-org/Qwen3-0.6B-GGUF ```

路西奥尔23B指令1

<a id="en"></a> **English** | 한국어 This model is a fine-tune of `Qwen/Qwen3-ASR-1.

This model provides a few variants of microsoft/Phi-4-mini-instruct that are ready for deployment on Android using the LiteRT (fka TFLite) stack, MediaPipe LLM Inference API (in maintenance-only...

GGUF / llama.