Open-Weight Models

Model · Speech recognition

wav2vec2-large-xlsr-53-chinese-zh-cn

Jonatas Grosman

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Chinese using the train and validation splits of Common Voice 6.1, CSS10 and ST-CMDS. When using this model, make sure that your speech input is sampled at 16kHz. This model has been fine-tuned thanks to the GPU credits generously given by the OVHcloud:) The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint The model can be used directly (without a language model) as follows... Using the HuggingSound library: The model can be evaluated as follows on the Chinese (zh-CN) test data of Common Voice. In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran…

Model · Speech recognition

wav2vec2-xls-r-300m-mixed

Mesolitica

Finetuned https://huggingface.co/facebook/wav2vec2-xls-r-300m on https://github.com/huseinzol05/malaya-speech/tree/master/data/mixed-stt This model was finetuned on 3 languages, This model trained on a single RTX 3090 Ti 24GB VRAM, provided by https://mesolitica.com/. Evaluation set from https://github.com/huseinzol05/malaya-speech/tree/master/pretrained-model/prepare-stt with sizes, It achieves the following results on the evaluation set based on evaluate-gpu.ipynb: Mixed evaluation, Malay evaluation, Singlish evaluation, Mandarin evaluation, Language model from https://huggingface.co/huseinzol05/language-model-bahasa-manglish-combined

Model

parrot_paraphraser_on_T5

Prithivi Da

Parrot is a paraphrase based utterance augmentation framework purpose built to accelerate training NLU models. A paraphrase framework is more than just a paraphrasing model. For more details on the library and usage please refer to the github page Huggingface lists 12 paraphrase models, RapidAPI lists 7 fremium and commercial paraphrasers like QuillBot, Rasa has discussed an experimental paraphraser for augmenting text data here, Sentence-transfomers offers a paraphrase mining utility and NLPAug offers word level augmentation with a PPDB (a multi-million paraphrase database). While these attempts at paraphrasing are great, there are still some gaps and paraphrasing is NOT yet a mainstream…

Model · Text generation

OpenELM-1_1B-Instruct

Apple

Sachin Mehta, Mohammad Hossein Sekhavat, Qingqing Cao, Maxwell Horton, Yanzi Jin, Chenfan Sun, Iman Mirzadeh, Mahyar Najibi, Dmitry Belenko, Peter Zatloukal, Mohammad Rastegari We introduce OpenELM, a family of Open Efficient Language Models. OpenELM uses a layer-wise scaling strategy to efficiently allocate parameters within each layer of the transformer model, leading to enhanced accuracy. We pretrained OpenELM models using the CoreNet library. We release both pretrained and instruction tuned models with 270M, 450M, 1.1B and 3B parameters. We release the complete framework, encompassing data preparation, training, fine-tuning, and evaluation procedures, alongside multiple pre-trained…

Model · Speech recognition

voice-activity-detection

Pyannote

Using this open-source model in production? Consider switching to pyannoteAI for better and faster options.

Model

flan-t5-base

Google

alt="drawing" width="600"/> If you already know T5, FLAN-T5 is just better at everything. For the same number of parameters, these models have been fine-tuned on more than 1000 additional tasks covering also more languages. As mentioned in the first few lines of the abstract: Disclaimer: Content from this model card has been written by the Hugging Face team, and parts of it were copy pasted from the T5 model card. Find below some example scripts on how to use the model in transformers: The authors write in the original paper's model card that: See the research paper for further details. The information below in this section are copied from the model's official model card: The model was…

Model · Speech recognition

Wav2Vec2-large-xlsr-hindi

Shyam Sunder Kumar

Fine-tuned facebook/wav2vec2-large-xlsr-53 hindi using the Multilingual and code-switching ASR challenges for low resource Indian languages. When using this model, make sure that your speech input is sampled at 16kHz. The model can be used directly (without a language model) as follows: The model can be evaluated as follows on the hindi test data of Common Voice. The script used for training can be found Hindi ASR Fine Tuning Wav2Vec2

Model · Sentence similarity

gte-multilingual-base

Alibaba-NLP

The gte-multilingual-base model is the latest in the GTE (General Text Embedding) family of models, featuring several key attributes: Paper: mGTE: Generalized Long-Context Text Representation and Reranking Models for Multilingual Text Retrieval - It is recommended to install xformers and enable unpadding for acceleration, refer to enable-unpadding-and-xformers. - How to use with TEI: refs/pr/7 Usage via docker and infinity, MIT Licensed. Usage via Docker and Text Embeddings Inference (TEI): Then you can send requests to the deployed API via the OpenAI-compatible v1/embeddings route (more information about the OpenAI Embeddings API): We validated the performance of the gte-multilingual-base…

Model

yolo-world-mirror

Dowon Hwang

https://docs.ultralytics.com/models/yolo-world/#available-models-supported-tasks-and-operating-modes model weights for ultralytics yolo models

Model · Tabular regression

mitra-regressor

Autogluon

Mitra regressor is a tabular foundation model that is pre-trained on purely synthetic datasets sampled from a mix of random regressors. Mitra is based on a 12-layer Transformer of 72 M parameters, pre-trained by incorporating an in-context learning paradigm. To use Mitra regressor, install AutoGluon by running: A minimal example showing how to perform inference using the Mitra regressor: This project is licensed under the Apache-2.0 License. Amazon Science blog: Mitra: Mixed synthetic priors for enhancing tabular foundation models

Model · Image and text to text

Qwen3.6-35B-A3B-NVFP4

Unsloth AI

1.56x faster throughput than other NVFP4 quants. This is an Unsloth NVFP4 quantized checkpoint calibrated on a mixture of our Unsloth dataset + UltraChat dataset. Works on a 32GB VRAM GPU. Benchmarks on 1xB200 128 concurrency. Use the 35B NVFP4 Fast version for 1.79x faster at a little less accuracy For accuracy benchmarks, we conducted MMLU-Pro, AIME 2025, GPQA for FP8, BF16, NVIDIA's NVFP4 and our NVFP4s - we show our faster quants do similarly on all: Read all benchmarks in our NVFP4 blog To install vLLM in a separate venv: Then to serve the 35B variant: You must use the below or you will get 2x slower inference! Also do NOT use the Marlin backend since it's 2x slower - use the native…

Model

KAT-Coder-V2.5-Dev-APEX-GGUF

Mudler

APEX (Adaptive Precision for EXpert Models) quantizations of Kwaipilot/KAT-Coder-V2.5-Dev — Kwaipilot's Mixture-of-Experts model for agentic coding. Brought to you by is a quantization strategy for Mixture-of-Experts (MoE) models. It classifies tensors by role (routed expert, shared expert, attention) and applies a layer-wise precision gradient — edge layers (first/last 5) get higher precision, middle layers compress more aggressively. I-variants use diverse imatrix calibration (chat, code, reasoning, tool-calling, agentic traces, Wikipedia). In MoE models the routed-expert FFN tensors dominate the weight budget but only ~8/256 experts activate per token, so APEX compresses middle-layer…

Model · Text generation

Qwen3.8-27B-OBLITERATED

OBLITERATUS

V3 applies iterative refinement on top of V2's complementary blend, with targeted corpus expansion. The result: genuine liberation — not just removal of hard refusals but elimination of safety-lecture deflections. - Genuinely answers restricted queries — provides real substance instead of safety lectures - 20/20 on code generation tasks — functional implementations, not disclaimers - Thinking ON compatible — no refusals in either thinking mode - Honest scoring — every response manually audited for real substance, not just absence of "I cannot" - -2.1pp MMLU — modest capability cost for genuine liberation If you're using this model in an agent harness (coding agent, pentest framework, etc.)…

Model · Speech recognition

wav2vec2-base-vi-vlsp2020

Binh Nguyen

Our models use wav2vec2 architecture, pre-trained on 13k hours of Vietnamese youtube audio (un-label data) and fine-tuned on 250 hours labeled of VLSP ASR dataset on 16kHz sampled speech audio. You can find more description here The ASR model parameters are made available for non-commercial use only, under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license. You can find details at: https://creativecommons.org/licenses/by-nc/4.0/legalcode [email protected]

Model · Speech recognition

wav2vec2-xls-r-300m-cs-250

Vojtěch Drábek

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the commonvoice 8.0 dataset as well as other datasets listed below. It achieves the following results on the evaluation set: The eval.py script results using a LM are: Fine-tuned facebook/wav2vec2-large-xlsr-53 on Czech using the Common Voice dataset. When using this model, make sure that your speech input is sampled at 16kHz. The model can be used directly (without a language model) as follows: The model can be evaluated using the attached eval.py script: The Common Voice 8.0 train and validation datasets were used for training, as well as the following datasets: - Šmídl, Luboš and Pražák, Aleš, 2013, OVM – Otázky…

Model · Image and text to text

Unlimited-OCR-AWQ

Sahil Chachra

AWQ 4-bit (W4A16) quantization of baidu/Unlimited-OCR, a 3B vision-language OCR model that pushes DeepSeek-OCR one step further (one-shot, long-horizon document parsing). This repo quantizes the DeepSeek-V2 MoE text decoder with activation-aware scaling (AWQ) while keeping the vision tower in BF16, so it stays a drop-in transformers model. Unlimited-OCR uses the DeepSeek-OCR prompt vocabulary. The prompt must contain; prefix it with whenever you also want bounding boxes for what was read. - base — basesize=1024, imagesize=1024, cropmode=False. Good default for normal pages. - gundam — basesize=1024, imagesize=640, cropmode=True. Tiles the page; use for dense or large/high-resolution…

Model · Text to speech

OmniVoice

K2 FSA

OmniVoice is a massively multilingual zero-shot text-to-speech (TTS) model supporting over 600 languages. Built on a novel diffusion language model-style architecture, it delivers high-quality speech with superior inference speed, supporting voice cloning and voice design. - 600+ Languages Supported: The broadest language coverage among zero-shot TTS models. To get started, install the omnivoice library: You can use OmniVoice for zero-shot voice cloning as follows: For more generation modes (e.g., voice design), functions (e.g., non-verbal symbols, pronunciation correction) and comprehensive usage instructions, see our GitHub Repository. You can directly discuss on GitHub Issues. You can…

Model · Image and text to text

Qwen3.8-27B-NVFP4

Inferact Inc.

Quantized version of https://huggingface.co/Qwen/Qwen3.8-27B

Model

flux2-dev

Comfy Org

Repackaged model files for ComfyUI. Place the files in the following folders: These are quantized weights. The original FLUX.2 repo is here: FLUX.2-dev. - Flux2TurboComfyv2.safetensors by ByteZSzn - Flux2-Turbo-LoRAcomfyui.safetensors by ByteZSzn

Model · Speech recognition

wav2vec2-large-xlsr-53-finnish

Jonatas Grosman

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Finnish using the train and validation splits of Common Voice 6.1 and CSS10. When using this model, make sure that your speech input is sampled at 16kHz. This model has been fine-tuned thanks to the GPU credits generously given by the OVHcloud:) The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint The model can be used directly (without a language model) as follows... Using the HuggingSound library: The model can be evaluated as follows on the Finnish test data of Common Voice. In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation…

Model · Summarization

bart-large-cnn

AI at Meta

BART model pre-trained on English language, and fine-tuned on CNN Daily Mail. It was introduced in the paper BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension by Lewis et al. and first released in this repository (https://github.com/pytorch/fairseq/tree/master/examples/bart). Disclaimer: The team releasing BART did not write a model card for this model so this model card has been written by the Hugging Face team. BART is a transformer encoder-encoder (seq2seq) model with a bidirectional (BERT-like) encoder and an autoregressive (GPT-like) decoder. BART is pre-trained by (1) corrupting text with an arbitrary noising function…

Model

mobilebert-uncased

Google

MobileBERT is a thin version of BERTLARGE, while equipped with bottleneck structures and a carefully designed balance between self-attentions and feed-forward networks. This checkpoint is the original MobileBert Optimized Uncased English: checkpoint.

Model · Zero shot image classification

CLIP-convnext_base_w-laion2B-s13B-b82K-augreg

LAION eV

A series of CLIP ConvNeXt-Base (w/ wide embed dim) models trained on subsets LAION-5B (https://laion.ai/blog/laion-5b/) using OpenCLIP (https://github.com/mlfoundations/openclip). Explore an alternative to ViT and ResNet (w/ AttentionPooling) CLIP models that scales well with model size and image resolution First known ConvNeXt CLIP models trained at scale in the range of CLIP ViT-B/16 and RN50x4 models First released model weights exploring increase of augmentation + regularization for image tower via adding (greater scale range of RRC, random erasing, stochastic depth) The models utilize the timm ConvNeXt-Base model (convnextbase) as the image tower, and the same text tower as the RN50x4…

Model · Zero shot image classification

siglip-so400m-patch14-384

Google

SigLIP model pre-trained on WebLi at resolution 384x384. It was introduced in the paper Sigmoid Loss for Language Image Pre-Training by Zhai et al. and first released in this repository. This model has the SoViT-400m architecture, which is the shape-optimized version as presented in Getting ViT in Shape: Scaling Laws for Compute-Optimal Model Design by Alabdulmohsin et al. Disclaimer: The team releasing SigLIP did not write a model card for this model so this model card has been written by the Hugging Face team. SigLIP is CLIP, a multimodal model, with a better loss function. The sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities…