Open-Weight Models

Model · Object detection

deformable-detr-DocLayNet

Aryn Inc.

Deformable DEtection TRansformer (DETR), trained on DocLayNet (including 80k annotated pages in 11 classes). You can use this model in the serverless Aryn Partitioning Service. You can get started here The DETR model is an encoder-decoder transformer with a convolutional backbone. Two heads are added on top of the decoder outputs in order to perform object detection: a linear layer for the class labels and a MLP (multi-layer perceptron) for the bounding boxes. The model uses so-called object queries to detect objects in an image. Each object query looks for a particular object in the image. For COCO, the number of object queries is set to 100. The model is trained using a "bipartite…

Model · Any to any

MiniCPM-o-4_5-awq

OpenBMB

A Gemini 2.5 Flash Level MLLM for Vision, Speech, and Full-Duplex Mulitmodal Live Streaming on | CaseBook(Audio, Omni Full-Duplex) MiniCPM-o 4.5 is the latest and most capable model in the MiniCPM-o series. The model is built in an end-to-end fashion based on SigLip2, Whisper-medium, CosyVoice2, and Qwen3-8B with a total of 9B parameters. It exhibits a significant performance improvement, and introduces new features for full-duplex multimodal live streaming. Notable features of MiniCPM-o 4.5 include: - Leading Visual Capability. MiniCPM-o 4.5 achieves an average score of 77.6 on OpenCompass, a comprehensive evaluation of 8 popular benchmarks. With only 9B parameters, it surpasses widely…

You will find here all the TTS models used with WanGP (https://github.com/deepbeepmeep/Wan2GP): WanGP by DeepBeepMeep: The best Open Source Video Generative Models Accessible to the GPU Poor WanGP supports the Wan (and derived models), Hunyuan Video, Flux 1 & 2, Qwen, Z-Image and LTV Video models with: Low VRAM requirements (as low as 6 GB of VRAM is sufficient for certain models) Support for old GPUs (RTX 10XX, 20xx,...) Very Fast on the latest GPUs Easy to use Full Web based interface Auto download of the required model adapted to your specific architecture Tools integrated to facilitate Video Generation: Mask Editor, Prompt Enhancer, Temporal and Spatial Generation Loras Support to…

https://github.com/vibevoice-community/VibeVoice VibeVoice is a novel framework designed for generating expressive, long-form, multi-speaker conversational audio, such as podcasts, from text. It addresses significant challenges in traditional Text-to-Speech (TTS) systems, particularly in scalability, speaker consistency, and natural turn-taking. A core innovation of VibeVoice is its use of continuous speech tokenizers (Acoustic and Semantic) operating at an ultra-low frame rate of 7.5 Hz. These tokenizers efficiently preserve audio fidelity while significantly boosting computational efficiency for processing long sequences. VibeVoice employs a next-token diffusion framework, leveraging a…

This is a LLaMA 3 Youko qlora fine-tune, created using a new version of the VNTL dataset. The purpose of this fine-tune is to improve performance of LLMs at translating Japanese visual novels to English. Unlike the previous version, this one doesn't includes the "chat mode". For this new version of VNTL 8B, I've rebuilt and expanded VNTL's dataset from the groud up, and I'm happy to say it performs really well, outperforming the previous version when it comes to accuracy and stability, it makes far fewer mistakes than it even when running at high temperatures (though I still recommend temperature 0 for the best accuracy). Some major changes in this version: - Switched to the default LLaMA3…

Model · Text to speech

Kokoro-82M-ONNX

ONNX Community

Kokoro is a frontier TTS model for its size of 82 million parameters (text in/audio out). First, install the kokoro-js library from NPM using: You can then generate speech as follows: Optionally, save the audio to a file: The model is resilient to quantization, enabling efficient high-quality speech synthesis at a fraction of the original model size.

Model · Summarization

text_summarization

Falcons.ai

The Fine-Tuned T5 Small is a variant of the T5 transformer model, designed for the task of text summarization. It is adapted and fine-tuned to generate concise and coherent summaries of input text. The model, named "t5-small," is pre-trained on a diverse corpus of text data, enabling it to capture essential information and generate meaningful summaries. Fine-tuning is conducted with careful attention to hyperparameter settings, including batch size and learning rate, to ensure optimal performance for text summarization. During the fine-tuning process, a batch size of 8 is chosen for efficient computation and learning. Additionally, a learning rate of 2e-5 is selected to balance convergence…

Model · Image segmentation

BiRefNet_HR

Peng Zheng

This repo is the official implementation of "Bilateral Reference for High-Resolution Dichotomous Image Segmentation" (CAAI AIR 2024). Visit our GitHub repo: https://github.com/ZhengPeng7/BiRefNet for more details -- codes, docs, and model zoo! This repo contains the weights of BiRefNet proposed in our paper, which has achieved the SOTA performance on three tasks (DIS, HRSOD, and COD). Go to my GitHub page for BiRefNet codes and the latest updates: https://github.com/ZhengPeng7/BiRefNet:) + Online Image Inference on Colab: + Online Inference with GUI on Hugging Face with adjustable resolutions: + Inference and evaluation of your given weights: + Many thanks to @freepik for their generous…

Model · Text to speech

speecht5_tts

Microsoft

SpeechT5 model fine-tuned for speech synthesis (text-to-speech) on LibriTTS. This model was introduced in SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing by Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, Furu Wei. SpeechT5 was first released in this repository, original weights. The license used is MIT. Motivated by the success of T5 (Text-To-Text Transfer Transformer) in pre-trained natural language processing models, we propose a unified-modal SpeechT5 framework that explores the encoder-decoder pre-training for self-supervised speech/text representation learning. The…

Model · Translation

opus-mt-tc-big-ar-en

Helsinki-NLP Research Group

Neural machine translation model for translating from Arabic (ar) to English (en). This model is part of the OPUS-MT project, an effort to make neural machine translation models widely available and accessible for many languages in the world. All models are originally trained using the amazing framework of Marian NMT, an efficient NMT implementation written in pure C++. The models have been converted to pyTorch using the transformers library by huggingface. Training data is taken from OPUS and training pipelines use the procedures of OPUS-MT-train. You can also use OPUS-MT models with the transformers pipelines, for example: The work is supported by the European Language Grid as pilot…

Model · Any to any

gemma-4-12B-it-FP8-dynamic

Thor Lin

Self-quantized FP8 (dynamic) of google/gemma-4-12B-it — Google's encoder-free omni model (text + image + audio + video). Quantized and benchmarked on an NVIDIA DGX Spark (GB10, sm121a). TL;DR: 13 GB on disk (from 23 GB BF16), 15.9 tok/s on a GB10 via vLLM, all four modalities intact. Data-free — no calibration needed. If you want the smallest + fastest build, see the sibling NVFP4 weight-only repo. FP8 is the conservative choice (dynamic activations, no calibration, widest kernel support). I scored all three formats on MMLU (English, 57 subjects) and TMMLU+ (Traditional Chinese, 66 subjects) with lm-evaluation-harness, 5-shot, chat template applied, limit=30 (N ≈ 1,710 EN / 1,980 TC, ±~1.0…

Model · Text to speech

Soprano-1.1-80M

Eugene Kwek

2026.01.14 - Soprano-1.1-80M released! 95% fewer hallucinations and a 63% preference rate over Soprano-80M. 2026.01.13 - Soprano-Factory released! You can now train/fine-tune your own Soprano models. 2025.12.22 - Soprano-80M released! Code | Demo Soprano is an ultra‑lightweight, on-device text‑to‑speech (TTS) model designed for expressive, high‑fidelity speech synthesis at unprecedented speed. Soprano was designed with the following features: - Up to 2000x real-time generation on GPU and 20x real-time on CPU - Lossless streaming with one plus one, etc) If Soprano produces unsatisfactory results, you can easily regenerate it for a new, potentially better generation. You may also change the…

Model · Zero-shot classification

nli-deberta-v3-large

Sentence Transformers - Cross-Encoders

This model was trained using SentenceTransformers Cross-Encoder class. This model is based on microsoft/deberta-v3-large The model was trained on the SNLI and MultiNLI datasets. For a given sentence pair, it will output three scores corresponding to the labels: contradiction, entailment, neutral. For futher evaluation results, see SBERT.net - Pretrained Cross-Encoder. Pre-trained models can be used like this: You can use the model also directly with Transformers library (without SentenceTransformers library): This model can also be used for zero-shot-classification

Model · Image to text

PP-OCRv5_mobile_det

PaddlePaddle

PP-OCRv5mobiledet is one of the PP-OCRv5det series, the latest generation of text detection models developed by the PaddleOCR team. It aims to efficiently and accurately supports the detection of text in diverse scenarios—including handwriting, vertical, rotated, and curved text—across multiple languages such as Simplified Chinese, Traditional Chinese, English, and Japanese. Key features include robust handling of complex layouts, varying text sizes, and challenging backgrounds, making it suitable for practical applications like document analysis, license plate recognition, and scene text detection. The key accuracy metrics are as follow: Please refer to the following commands to install…

Model · Text to speech

MeloTTS-Chinese

MyShell.ai

MeloTTS is a high-quality multi-lingual text-to-speech library by MyShell.ai. Supported languages include: - The Chinese speaker supports mixed Chinese and English. - Fast enough for CPU real-time inference. An unofficial live demo is hosted on Hugging Face Spaces. There are hundreds of TTS models on MyShell, much more than MeloTTS. See examples here. More can be found at the widget center of MyShell.ai. Follow the installation steps here before using the following snippet: Open Source AI Grant We are actively sponsoring open-source AI projects. The sponsorship includes GPU resources, fundings and intellectual support (collaboration with top research labs). We welcome both reseach and…

Model · Text classification

DeBERTa-v3-base-mnli-fever-docnli-ling-2c

Moritz Borrett-Laurer (formerly Laurer)

This model was trained on 1.279.665 hypothesis-premise pairs from 8 NLI datasets: MultiNLI, Fever-NLI, LingNLI and DocNLI (which includes ANLI, QNLI, DUC, CNN/DailyMail, Curation). It is the only model in the model hub trained on 8 NLI datasets, including DocNLI with very long texts to learn long range reasoning. Note that the model was trained on binary NLI to predict either "entailment" or "not-entailment". The DocNLI merges the classes "neural" and "contradiction" into "not-entailment" to enable the inclusion of the DocNLI dataset. The base model is DeBERTa-v3-base from Microsoft. The v3 variant of DeBERTa substantially outperforms previous versions of the model by including a different…

Model · Translation

opus-mt-da-en

Helsinki-NLP Research Group

source languages: da; target languages: en; OPUS readme: da-en; dataset: opus; model: transformer-align; pre-processing: normalization + SentencePiece.

Model · Text to image

girlslike-zimage

Ifmylove2011

用于测试基于 Z-Image Base/Turbo BF16 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。 后缀为zi代表基于z-image base训练，zit则代表基于turbo训练，zi在base与turbo模型中均可使用，zit则只能在turbo中使用。 z-image base训练尚无较为通用的可行方案，目前本zi系列选择的方案是全秩lokr，可在与turbo模型配合使用时以权重1.0出图，也能适当增加权重以增强相似性，最好不要超过1.5。 而在与base模型或相关微调模型配合使用时，可适当降低权重。 基于base训练的zi lokr在turbo模型中使用时，质感会更接近turbo本身（应该是base模型本身低噪部分难以被"污染"的缘故），中远景图中的人物相似度也会比zit好一些，近景则更"平淡"。 但总体来说兼容性存疑，比如解剖学问题仍然大量存在（在base模型上生图就存在类似问题，也影响到了训练，会让肢体问题重新变得不稳定又棘手），其他类似的问题都还在测试中，后续有更好的方案会继续更新。 z-image turbo bf16； ersde + sgmuniform | euler + flowmatch； cfg=1； lora权重0.5~1.5； 无其他lora参与。 鉴于z-image…

This model is a core component of the meikiocr pipeline. For the full implementation, command-line script, and documentation, please see the official GitHub repository. meiki.text.detect.v0.1 is an update to meiki.text.detect.v0 (see below): - meiki.text.detect.v0.1 is a new state-of-the-art, open weight text detection model for video games beating text detection models like PaddleOCR - while it is still based on D-FINE detector, it uses mobilenet v4 small as backbone instead of hgnet v2 - v0.1 models increase focus on video game text detection and are limited to 64 detected boxes, increasing efficency for this use case (making them less suitable for manga text detection out of the box)…

Model · Zero-shot classification

distilbart-mnli-12-3

Suraj Patil

distilbart-mnli is the distilled version of bart-large-mnli created using the No Teacher Distillation technique proposed for BART summarisation by Huggingface, here. We just copy alternating layers from bart-large-mnli and finetune more on the same data. This is a very simple and effective technique, as we can see the performance drop is very little. Detailed performace trade-offs will be posted in this sheet. If you want to train these models yourself, clone the distillbart-mnli repo and follow the steps below Clone and install transformers from source Download MNLI data Create student model Start fine-tuning You can find the logs of these trained models in this wandb project.

Unmodified 4-step and 8-step DaSiWa MiniMax H3 Hybrid SafeTensors checkpoints mirrored for BRP Canvas downloads. The hybrid checkpoint supports text-to-video, reference-to-video, and first/last-frame-to-video through the same ComfyUI workflow. This is not an official MiniMax or DaSiWa distribution. Review the original model page and the included upstream MiniMax license before use. BRP Canvas defaults to shift video 9 and shift audio 4 for both distilled checkpoints. - 4-step: 56c52c7890c105308d28fe9c25c25fdb80e6cd6a54e2604d8af71732ba4ed74f - 8-step: e0441d26414f6e0c28f43d580e6cc56fad424da0fa4d261b698ca73188aa6332

Model · Zero-shot classification

multilingual-MiniLMv2-L12-mnli-xnli

Moritz Borrett-Laurer (formerly Laurer)

multilingual - zero-shot-classification - text-classification - nli - pytorch - accuracy - multinli - xnli pipelinetag: zero-shot-classification candidatelabels: "politics, economy, entertainment, environment" This multilingual model can perform natural language inference (NLI) on 100+ languages and is therefore also suitable for multilingual zero-shot classification. The underlying multilingual-MiniLM-L12 model was created by Microsoft and was distilled from XLM-RoBERTa-large (see details in the original paper and newer information in this repo). The model was then fine-tuned on the XNLI dataset, which contains hypothesis-premise pairs from 15 languages, as well as the English MNLI…

Model · Any to any

gemma-4-12B-it-qat-AWQ-INT4

Cyankiwi

Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context window of up to 256K tokens and maintains multilingual support in over 140 languages. Featuring both Dense and Mixture-of-Experts (MoE) architectures, Gemma 4 is well-suited for tasks like text generation, coding, and reasoning. The models are available in five distinct sizes: E2B, E4B, 12B, 26B A4B, and 31B. Their diverse sizes make them deployable in environments ranging from…

Model · Any to any

gemma-4-12B-it-GGUF

GGML Org

Run with https://llama.app - https://huggingface.co/google/gemma-4-12B-it - https://huggingface.co/google/gemma-4-12B-it-assistant - https://huggingface.co/google/gemma-4-12B-it-qat-q40-unquantized-assistant - https://huggingface.co/google/gemma-4-12B-it-qat-q40-unquantized - add info - add dflash