Open-Weight Models

Model · Time series forecasting

timesfm-2.0-500m-pytorch

Google

TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting. This is not an officially supported Google product. timesfm-2.0-500m is the second open model checkpoint: - It performs univariate time series forecasting for context lengths up to 2048 time points and any horizon lengths, with an optional frequency indicator. Note that it can go even beyond 2048 context even though it was trained with that as the maximum context. - It focuses on point forecasts. We experimentally offer 10 quantile heads but they have not been calibrated after pretraining. - It ideally requires the context to be contiguous (i.e. no…

Model · Text to speech

Kokoro-82M-bf16

MLX Community

This model was converted to MLX format from hexagrad/Kokoro-82M using mlx-audio version 0.0.1. Refer to the original model card for more details on the model.

This is an image captioning model trained by @ydshieh in flax this is pytorch version of this. https://ankur3107.github.io/blogs/the-illustrated-image-captioning-using-transformers/ https://huggingface.co/ankur310794 https://twitter.com/ankur310794 http://github.com/ankur3107 https://www.linkedin.com/in/ankur310794

Model · Image segmentation

oneformer_ade20k_swin_tiny

SHI Labs

OneFormer model trained on the ADE20k dataset (tiny-sized version, Swin backbone). It was introduced in the paper OneFormer: One Transformer to Rule Universal Image Segmentation by Jain et al. and first released in this repository. OneFormer is the first multi-task universal image segmentation framework. It needs to be trained only once with a single universal architecture, a single model, and on a single dataset, to outperform existing specialized models across semantic, instance, and panoptic segmentation tasks. OneFormer uses a task token to condition the model on the task in focus, making the architecture task-guided for training, and task-dynamic for inference, all with a single model.…

Model · Text to video

LTX-2.5-Distilled-GGUF

Jay

This repository provides quantized GGUF formats of the distilled transformer from Lightricks/LTX-2.5. These weights are highly optimized for local execution, allowing you to run high-fidelity video and audio generation workflows on hardware with memory constraints while retaining the core visual fidelity of the original base model. LTX-2.5 operates on a split-component architecture. To run these GGUF diffusion models in environments like ComfyUI or local Python pipelines, you must also fetch the official Text Encoders and VAEs directly from the upstream Lightricks/LTX-2.5 repository. Download these from the textencoders directory: gemma4-12b-with-proj-ltx-2.5-bf16.safetensors (26.3 GB)…

Model · Text to video

Wan2.2-TI2V-5B-GGUF

Unsloth AI

Every GGUF quantisation of Wan2.2-TI2V-5B that QuantStack/Wan2.2-TI2V-5B-GGUF publishes, plus the companion VAE, mirrored here. Unsloth Studio offers this repo as the curated one-click GGUF pick for Wan2.2 TI2V 5B, so its availability is Studio's problem rather than the repacker's: a rename or a takedown turns the pick into a 404 no client can work around. All 13 quants are mirrored, not a chosen few, because the picker lets you choose the precision. The weights are unmodified: byte for byte the files of the same name in the source repo. TI2V-5B is a 720P-only checkpoint: the supported sizes are 1280x704 and 704x1280, and its VAE has temporal compression 4, so valid frame counts are 4k+1.…

Model · Text to image

diving-illustrious-real-asian-v50-sdxl

John Smith

Original model is here. This model created by DivingSuit.

Model · Audio classification

mms-lid-126

AI at Meta

This checkpoint is a model fine-tuned for speech language identification (LID) and part of Facebook's Massive Multilingual Speech project. This checkpoint is based on the Wav2Vec2 architecture and classifies raw audio input to a probability distribution over 126 output classes (each class representing a language). The checkpoint consists of 1 billion parameters and has been fine-tuned from facebook/mms-1b on 126 languages. This MMS checkpoint can be used with Transformers to identify the spoken language of an audio. It can recognize the following 126 languages. Let's look at a simple example. First, we install transformers and some other libraries Note: In order to use MMS you need to have…

Model · Text to speech

MOSS-TTS-Nano-100M

OpenMOSS

MOSS-TTS-Nano is an open-source multilingual tiny speech generation model from MOSI.AI and the OpenMOSS team. With only 0.1B parameters, it is designed for realtime speech generation, can run directly on CPU without a GPU, and keeps the deployment stack simple enough for local demos, web serving, and lightweight product integration. 2026.4.10: We release MOSS-TTS-Nano. A demo Space is available at OpenMOSS-Team/MOSS-TTS-Nano. You can also view the demo and more details at openmoss.github.io/MOSS-TTS-Nano-Demo/. MOSS-TTS-Nano focuses on the part of TTS deployment that matters most in practice: small footprint, low latency, good enough quality for realtime products, and simple local setup. It…

Model · Object detection

Anzhcs_YOLOs

Anzhc

YOLOs in this repo are trained with datasets that i have annotated myself, or with the help of my friends(They will be appropriately mentioned in those cases). YOLOs on open datasets will have their own pages. Ultralytics 8.3.217 updates mask handling, which breaks function in main Adetailer repo. Install Ultralytics==8.3.216 or lower. Alternatively - use forks that fix this. - Fixed in main repo. I've added some features to make Adetailer more usable and less manual - https://github.com/Anzhc/aadetailer-reforge Im open to commissions, hit me up in Discord - anzhc P.S. All model names in tables have download links attached:3 Series of models aiming at detecting and segmenting face…

Donut model pre-trained-only. It was introduced in the paper OCR-free Document Understanding Transformer by Geewok et al. and first released in this repository. Disclaimer: The team releasing Donut did not write a model card for this model so this model card has been written by the Hugging Face team. Donut consists of a vision encoder (Swin Transformer) and a text decoder (BART). Given an image, the encoder first encodes the image into a tensor of embeddings (of shape batchsize, seqlen, hiddensize), after which the decoder autoregressively generates text, conditioned on the encoding of the encoder. This model is meant to be fine-tuned on a downstream task, like document image classification…

Model · Text to speech

svara-tts-v1

Kenpath Labs

svara-TTS is a developer-first multilingual TTS model for 19 languages (18 Indic + Indian English). Built on an Orpheus-style discrete audio token approach, it targets clarity, expressiveness, and low-latency on commodity GPUs/CPUs. It supports light-weight emotion/style control (e.g.,,,, ) and simple speaker identities (Language (Gender)), with zero-shot adaptation paths. Try it live on the Demo Space, or on Colab Deployment scripts and inference repo will be available soon. Watch our Github for updates - Place style/emotion tags at the end of the sentence: आज... सच में अच्छी खबर है — शाम को मिलते हैं! - Use punctuation to hint prosody (ellipses, commas, exclamation). - For technical or…

Model · Translation

mbart-large-50-many-to-many-mmt

AI at Meta

This model is a fine-tuned checkpoint of mBART-large-50. mbart-large-50-many-to-many-mmt is fine-tuned for multilingual machine translation. It was introduced in Multilingual Translation with Extensible Multilingual Pretraining and Finetuning paper. The model can translate directly between any pair of 50 languages. To translate into a target language, the target language id is forced as the first generated token. To force the target language id as the first generated token, pass the forcedbostokenid parameter to the generate method. See the model hub to look for more fine-tuned versions. Arabic (arAR), Czech (csCZ), German (deDE), English (enXX), Spanish (esXX), Estonian (etEE), Finnish…

Model · Text to image

Krea2_turbo_diff_loras

Constantinez

These LoRAs were extracted from fine-tuned checkpoints. I did this primarily for myself because LoRAs are easier to work with: they can be mixed in real time and assigned different weights/strengths. This also saves space on your local computer. They work well at 100–150% strength. Choose a rank according to your taste and your hardware capabilities (the lower the rank, the less memory you need). However, some of my test generations showed that a lower rank can sometimes be preferable: it does not cause artifacts, and the result is closer to the reference image generated using the checkpoint from which the LoRA was extracted. A lot of time, electricity, and compute went into this on my…

Model · Text to image

Krea2_FP8

Alper

This repository provides an optimized FP8 (float8e4m3fn) weight-only quantized version of the newly released Krea 2 OSS (Turbo) transformer. This optimization reduces the model size from the original 24.76 GiB (BF16) down to 12.01 GiB, making it highly accessible and runnable on standard consumer hardware (such as 16GB and 24GB GPUs) without sacrificing output quality. Unlike generic global quantization scripts that aggressively convert every parameter (which often degrades generation details or introduces NaN/promotion calculation errors in neural networks), this model was quantized using a selective weight-only strategy: 1. Targeted Quantization: Only 2D floating-point weight matrices…

Model · Object detection

yolov11-license-plate-detection

Nuttapong Chimwai

This is a fine-tuned version of YOLOv11 (n, s, m, l, x) specialized for License Plate Detection, using a public dataset from Roboflow Universe: The upstream Roboflow dataset (license-plate-recognition-rxg4e) contains train/test contamination — the same source images appear in both the training and test splits with only minor manual augmentation applied (see Discussion #2 for concrete examples). As a result: - The reported metrics below are likely overestimated, because the test set is not a true held-out evaluation. - Real-world generalization performance is expected to be lower than the numbers in the table. - Treat all evaluation figures with caution and validate the model on your own…

Model · Text to image

nunchaku-qwen-image-edit

Nunchaku

This repository contains Nunchaku-quantized versions of Qwen-Image-Edit, an image-editing model based on Qwen-Image, advances in complex text rendering. It is optimized for efficient inference while maintaining minimal loss in performance. No recent news. Stay tuned for updates! Data Type: INT4 for non-Blackwell GPUs (pre-50-series), NVFP4 for Blackwell GPUs (50-series). Rank: r32 for faster inference, r128 for better quality but slower inference. Standard inference speed models for general use 4-step distilled models fused with Qwen-Image-Edit-Lightning-4steps-V1.0 LoRA using LoRA strength = 1.0 8-step distilled models fused with Qwen-Image-Edit-Lightning-8steps-V1.0 LoRA using LoRA…

HuggingFace mirror of https://civitai.com/models/1307155 You should use nsfwsks to trigger the image generation. Download them in the Files & versions tab.

Model · Text to image

FLUX.2-klein-4B-mflux-4bit

Runpod

4-bit quantized weights of [FLUX.2 [klein] 4B](https://huggingface.co/black-forest-labs/FLUX.2-klein-4B) by Black Forest Labs, optimized for mflux on Apple Silicon. FLUX.2 [klein] 4B is a 4 billion parameter rectified flow transformer by Black Forest Labs for fast image generation and editing. It delivers state-of-the-art quality with sub-second inference on consumer hardware. - Apache 2.0 — fully open for commercial use Apache 2.0, inherited from the original model.

Model · Image to text

PP-OCRv6_medium_det

PaddlePaddle

PP-OCRv6: From 1.5M to 34.5M Parameters, Surpassing Billion-Scale VLMs on OCR Tasks PP-OCRv6 is a lightweight OCR system that combines architectural innovation with data-centric optimization. It redesigns the backbone, detection neck, and recognition neck around a unified MetaFormer-style building block with structural reparameterization. Three model tiers (medium, small, tiny) share the same block primitives, covering deployment scenarios from server to edge. 1. Unified and Scalable Model Family: A three-tier OCR model family spanning 1.5M to 34.5M parameters. PP-OCRv6medium achieves 86.2% detection Hmean and 83.2% recognition accuracy, outperforming PP-OCRv5server by +4.6% and +5.1%…

Model · Text to image

Illustrious-xl-early-release-v0

OnomaAI

/ Title Base Styling / rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.8.2/css/all.min.css" Illustrious XL is the Illustration focused Stable Diffusion XL model which is continued from Kohaku XL Beta 5, trained by OnomaAI Research Team. The model focuses on utilizing large-scale annotated dataset, Danbooru2023. We release the v0.1 and v0.1-GUIDED model here, under fair public ai license, however discourages the usage of model over monetization purpose / any closed source purposes. For full technical details, please refer to our technical report. We plan to release several aesthetic-finetuned model variants in near future. By using this model, users agree to…

Model · Text to image

Krea-2-Raw

KREA

This is the Krea 2 Raw checkpoint, its not recommended for inference use. Its a good base for finetuning or post-training for your own needs and domains. For example, one use-case is to train LoRAs on midtrain and directly use them on Krea 2 Turbo. See our collection of in-house trained LoRAs trained on Raw and meant to be used with Turbo: Krea-2 LoRA Collection 1. Setup the official Krea 2 codebase 2. Download raw.safetensors in this repo 3. export OSSRAW= Install diffusers from source (for Krea2Pipeline): Install SGLang from source (https://github.com/sgl-project/sglang) See the full SGLang Krea 2 Cookbook here This model card covers the Krea 2 model family, including the following…

Model · Text to speech

MOSS-TTS-Local-Transformer-v1.5

OpenMOSS

MOSS-TTS-Local-Transformer-v1.5 is continued from MOSS-TTS-Local-Transformer-v1.0. It preserves the main 1.0 capabilities, including zero-shot voice cloning, long-form speech generation, token-level duration control, Pinyin/IPA pronunciation control, multilingual synthesis, and code-switching. For the full 1.0 feature walkthrough, input schema, and evaluation tables, please refer to the MOSS-TTS-Local-Transformer-v1.0 README. Compared with MOSS-TTS-Local-Transformer-v1.0, v1.5 focuses on the following improvements: - Stronger multilingual synthesis with language tags: when the language field is omitted, v1.5 may improve some languages and regress slightly on others compared with 1.0. When…

Model · Text to image

girlslike-krea2

Ifmylove2011

用于测试基于krea2raw int8训练的人物 LoKR / LoRA，仅代表个人审美与训练效果。实际生成结果仅供测试，请勿用于冒充、欺骗或其他不当用途。 理论上在krea2 raw模型与krea2 turbo模型上均可使用，使用强度在0.8~1.5之间，高于2.0面部会开始出现明显变形； （训练方案经过一段时间的测试，仍然采用了全秩lokr，泛化、细节、相似度三者相对均衡，在训练素材质量不高的情况下也更加有效） （存疑，某些中文概念确实可以准确理解，但也不要因此对krea2的中文理解能力抱有过高的期待，目前只能说比之前的flux系列有较大进步，实际效果有待商榷，比如flux系列的老问题——krea2仍然会把中文的"桃子"大概率画成苹果） krea2 turbo fp8； euler + beta|beta57； cfg=1； lora权重1.0； 无其他lora参与。 本系列lora训练中同样添加了人物本名作为触发词，生图时需输入触发词才能准确画出对应人物（名字本身就是触发词，无需其他前缀后缀）； 触发词一般为中文，部分人物将使用英文触发词，英文触发词会包含在示例图的水印中（水印通常以girlslikekrea2 for xxx形式出现，xxx即为触发词），有水印的示例图就代表需要使用英文触发词（因为krea2目前仍然无法准确生成中文，所以水印无法使用中文来表示触发词）； 示例图片一如既往包含工作流与提示词，lora强度为1.0，均由krea2turbo fp8模型生成； 示例图人名缩写可在 girlslikeloragalleryapp…