weighted/imatrix quants of https://huggingface.co/erax-ai/EraX-VL-7B-V2.0-Preview For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/EraX-VL-7B-V2.0-Preview-GGUF This is a vision model - mmproj files (if any) will be in the static repository. If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here are Artefact2's thoughts on…
SAVRN Model Hub
Open-Weight Models
An open-weight model is an AI model whose trained weights are published for anyone to download. The weights are what the model learned in training. With a copy of them you can run the model on hardware you control and train it further on your own data.
Open weights are not the same as open source. Many publishers release the weights without the training data or code, and the license sets what you may do with the model. This library puts each model's full card, architecture, files, license and published evaluations on one page.
2,760Models
859Datasets
254Papers
1,692Publishers
5,040Sourced relationships
Updated 2026-09-18 · How the library is built
2,760 models, sorted by most downloaded.
For usage instructions follow openai/whisper-large-v3-turbo. Note for faster-whisper vocab changes make model.ismultilingual and suppresstokens wrong. Please adjust the code as required if you want to use this with faster-whisper. Turbo finetune with japanese tokenizer. Full finetune trained 2^19 steps, batch size 64. Smaller vocab with ~1.6x bytes/token allows faster speed with 4 layers vs 2 layer distil (10% larger decoder). Benchmarks. Short form slightly behind v0.2 (trained less?) but long form much better. Also trained for lyrics but untested. Research supported with Cloud TPUs from Google's TPU Research Cloud (TRC)
X-CLIP model (base-sized, patch resolution of 16) trained fully-supervised on Kinetics-400. It was introduced in the paper Expanding Language-Image Pretrained Models for General Video Recognition by Ni et al. and first released in this repository. This model was trained using 16 frames per video, at a resolution of 224x224. Disclaimer: The team releasing X-CLIP did not write a model card for this model so this model card has been written by the Hugging Face team. X-CLIP is a minimal extension of CLIP for general video-language understanding. The model is trained in a contrastive way on (video, text) pairs. This allows the model to be used for tasks like zero-shot, few-shot or fully…
Haystack is an AI orchestration framework to build customizable, production-ready LLM applications. You can use this model in Haystack to do extractive question answering on documents. To load and run the model with Haystack: For a complete example with an extractive question answering pipeline that scales over many documents, check out the corresponding Haystack tutorial. - Timo Möller: timo.moeller [at] deepset.ai deepset is the company behind the production-ready open-source AI framework Haystack. We also have a Discord community open to everyone!
VideoMAE model(large) variant that has been finetuned for multi-label video classification (a video can belong to multiple classes simultaneously) for camera motion classification on internal dataset. The model predicts 18 different camera motion 'arcleft', 'arcright', 'dollyin', 'dollyout', 'panleft', 'panright', 'pedestaldown', 'pedestalup', 'rollleft', 'rollright', 'static', 'tiltdown', 'tiltup', 'truckleft', 'truckright', 'undefined', 'zoomin', 'zoomout' and and 3 shot type classes: 'pov', 'shake', 'track'. Model was trained to associate entire video with camera labels, not frame-level motions(!): [input video] -> label/labels (because multilabel) for all video. So, if this camera…
A frontier video understanding model developed by FAIR, Meta, which extends the pretraining objectives of VJEPA, resulting in state-of-the-art video understanding capabilities, leveraging data and model sizes at scale. The code is released in this repository. This is V-JEPA 2 ViT-g 384 model with video classification head pretrained on Something-Something-V2 dataset. To run V-JEPA 2 model, ensure you have installed the latest transformers
weighted/imatrix quants of https://huggingface.co/TurboAiLabs/turbo-ai-7b For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/turbo-ai-7b-GGUF If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here are Artefact2's thoughts on the matter: https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9 See…
This model is the fine-tuned version of "RoBERTa Base" using CUAD dataset https://huggingface.co/datasets/cuad For the use of the model with CUAD: https://github.com/marshmellow77/cuad-demo and https://huggingface.co/spaces/akdeniz27/contract-understanding-atticus-dataset-demo
bert-small-cord19-squad model fine-tuned on the CORD-19 QA dataset. The CORD-19 QA dataset is a SQuAD 2.0 formatted list of question, context, answer combinations covering the CORD-19 dataset.
https://huggingface.co/distilbert-base-cased-distilled-squad with ONNX weights to be compatible with Transformers.js. If you haven't already, you can install the Transformers.js JavaScript library from NPM using: Note: Having a separate repo for ONNX weights is intended to be a temporary solution until WebML gains more traction. If you would like to make your models web-ready, we recommend converting to ONNX using Optimum and structuring your repo like this one (with ONNX weights located in a subfolder named onnx).
plT5 models are T5-based language models trained on Polish corpora. The models were optimized for the original T5 denoising target. plT5 was trained on six different corpora available for Polish language: The training dataset was tokenized into subwords using a sentencepiece unigram model with vocabulary size of 50k tokens. If you use this model, please cite the following paper: The model was trained by Machine Learning Research Team at Allegro and Linguistic Engineering Group at Institute of Computer Science, Polish Academy of Sciences. You can contact us at: [email protected]
PTT5 Summ is a fine-tuned PTT5 model to perform Abstractive Summarization in Brazilian Portuguese texts. This model was fine-tuned on the datasets: RecognaSumm, WikiLingua, XL-Sum, TeMário.pdf) and CSTNews. For further information, please go to PTT5 Summ repository. author="Paiola, Pedro H. and de Rosa, Gustavo H. and Papa, Jo{\~a}o P.", editor="Xavier-Junior, Jo{\~a}o Carlos and Rios, Ricardo Ara{\'u}jo", title="Deep Learning-Based Abstractive Summarization for Brazilian Portuguese Texts", booktitle="BRACIS 2022: Intelligent Systems", year="2022", publisher="Springer International Publishing", address="Cham", pages="479--493", isbn="978-3-031-21689-3"} This model was fine-tuned using the…
GGUF quantizations of MohamedAhmedAE/Llama-3.2-3B-Instruct-Medical-Finetuned-merged — a 3.21B-parameter Llama-3.2 model adapted to the medical domain using QLoRA, with the LoRA weights merged back into the base. Quantized by tinyopsec. Med-LLaMA3.2-3B is the 3B (balanced / mid-tier) member of the Med-LLaMA3 family introduced in the paper "Med-LLaMA3: Advancing Medical Question-Answering Through Parameter-Efficient Fine-Tuning of Large Language Models" (Applied Sciences, 2026). The family adapts the LLaMA-3 architecture to the medical domain by training only a small fraction of the base model's parameters (5.70% for this 3B variant), achieving strong medical question-answering performance…
This repository contains a text-only Mini-K3-1H v2 pretraining checkpoint from a controlled 20-architecture comparison. The family retains Kimi-K3's KDA and Gated MLA operators, block Attention Residuals, Stable LatentMoE, SiTU activations, output gates, and Quantile Balancing at approximately one billion logical parameters. The exact architecture for this repository is listed below; some ablations deliberately replace the baseline KDA/MLA ratio, decay granularity, convolution length, or positional encoding. - Hidden width / attention heads / KDA head width: 1024 / 12 / 128 - Vocabulary / BOS / generation EOS / PAD: 163840 / 163584 / 163586 / 163839 control state retained in FP32 where…
GGUF quantizations of DualMind for local inference via llama.cpp, Ollama, LM Studio, and other GGUF-compatible runtimes. DualMind is a 1.7B parameter model that implements a dual-cognition reasoning architecture: The model learns to reason freely, then critique its own reasoning, then produce a final answer. Multi-model dialectics collapsed into shared weights. Training lineage: Qwen3-1.7B → DistilQwen3 (uncensored) → Disctil (DISC-refined) → TKD from Qwen3-30B-A3B-Thinking → DualMind SFT on LogicInferenceOA dataset. - temperature: 0.6 - topp: 0.9 - repeatpenalty: 1.3 (important — prevents enumeration loops) - numpredict: 512–1024 - DualMind — source model (SafeTensors)…
This is the 8B (high-capacity flagship) member of the Med-LLaMA3 family introduced in the paper “Med-LLaMA3: Advancing Medical Question-Answering Through Parameter-Efficient Fine-Tuning of Large Language Models” (Applied Sciences, 2026). The family adapts the LLaMA-3 architecture to the medical domain by training only a small fraction of the base model’s parameters (4.01% for this 8B variant), achieving strong medical question-answering performance while keeping the memory footprint low — enabling development and inference on low-cost, consumer-grade hardware. The 8B variant is the high-capacity model for complex clinical reasoning. It attains a mean accuracy of 75.71% across the eight MMLU…
C/C++ — https://github.com/ggml-org/llama.cpp
This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on the TradeNewsSum dataset for multilingual abstractive summarization of foreign trade news in Russian and English. This is a multilingual summarization model trained on economic and foreign trade news in Russian and English. It is based on the facebook/mbart-large-50-many-to-many-mmt architecture and fine-tuned specifically for the task of generating concise, informative summaries for news articles in the domain of international trade. - Generating abstractive summaries of foreign trade-related news in Russian and English. - Assisting analysts and journalists working with economic content. - Could be integrated…
GGUF quantizations of DualMinded-Qwen3-1.7B for local inference via llama.cpp, Ollama, LM Studio, and other GGUF-compatible runtimes. DualMinded-Qwen3-1.7B is the Opus-trained variant of the DualMind architecture. While DualMind was trained on LogicInferenceOA, DualMinded was trained on Opus-4.6-Reasoning-3000x-filtered — high-quality reasoning traces from Claude Opus 4.6. The Opus training data provides longer, more structured reasoning chains. The thinking column maps directly to the phase without heuristic splitting, producing cleaner cognitive transitions. Training lineage: Qwen3-1.7B → DistilQwen3 → Disctil → TKD checkpoint-512 → DualMind SFT v2 on Opus-4.6-Reasoning. Both share the…
Original TF 1 code here Authors: Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu on Dec 18, 2019 The following is copied from the authors' README. We train a pegasus model with sampled gap sentence ratios on both C4 and HugeNews, and stochastically sample important sentences. The updated the results are reported in this table. The "Mixed & Stochastic" model has the following changes: - trained on both C4 and HugeNews (dataset mixture is weighted by their number of examples). - trained for 1.5M instead of 500k (we observe slower convergence on pretraining perplexity). - the model uniformly sample a gap sentence ratio between 15% and 45%. - importance sentences are sampled using a…
This model is fine-tuned to generate summaries based on the input provided. It has been fine-tuned on a wide range of Persian news data, including BBC news and pnsummary. If you find this model useful, make a link to the huggingface model.
weighted/imatrix quants of https://huggingface.co/thelamapi/next-12b For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/next-12b-GGUF This is a vision model - mmproj files (if any) will be in the static repository. If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here are Artefact2's thoughts on the matter…
Model Collections
Hand-picked starting points, each with the reason it exists.
Collection · 4 entries
Embedding models for retrieval
Sentence and document embedding models used to build retrieval systems. Dimension and sequence length matter more than size here, and both come from the publisher.
Collection · 4 entries
Models that fit on one accelerator
Models whose publisher-reported parameter count puts them within reach of a single accelerator at common precisions. Memory needed depends on precision and serving configuration, so treat the parameter count as the starting point, not the answer.
Collection · 6 entries
Open-weight text models worth knowing
Widely used open-weight language models, chosen because each one is a distinct family rather than a variant of the one above it. Selection, not a ranking.
Collection · 3 entries
Speech and audio models
Recognition and synthesis models, grouped so the two directions are easy to compare.
Open-Weight Models Explained
What is an open-weight model?
An AI model whose trained weights are published for anyone to download, so it can be run, tested and fine-tuned on hardware the user controls.
Is an open-weight model the same as open source?
Not always. Open weights means the trained model can be downloaded. Open source usually also means the training code and data are available and the license allows broad reuse. Many open-weight models release the weights only.
Can I use an open-weight model commercially?
It depends on the license. Apache 2.0 and MIT allow commercial use. Other licenses limit it, for example to non-commercial use or below a set number of users. Every model page here shows its license.
How much memory does an open-weight model need?
About two bytes per parameter at 16-bit precision, so a 7-billion-parameter model needs roughly 14 GB for its weights, plus memory for the context it processes. Each model page lists its parameter count and the size of its files.
Related SAVRN Research
The hub sits beside SAVRN's market data and infrastructure research: what models cost to run, and what it takes to run them.
SAVRN Index
What open models cost to run
The same open-weight model priced by every host that serves it, per million tokens.
Research Hub
Data center trackers and maps
Moratoriums, permits, power, water and capital behind the facilities that run these models.
Method
How the Model Hub is built
Sources, evidence labels, refresh behaviour, and the limits of every comparison here.