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.
Updated 2026-10-18 · How the library is built
11,122 models, sorted by most downloaded.
Model files for a private video worker (Wan 2.2 image-to-video), loaded through RunPod's model cache. Includes an adult-content add-on (LoRA), so this collection is marked not for all audiences.
This qwen35 model was trained 2x faster with Unsloth and Huggingface's TRL library.
Cuantizado por el worker Cypher v3.5.1 desde Dennis1315/ward-web-14b (safetensors fp16 limpios). - Q4KM: model-q4km.gguf (5.78 GB) Q4KM = punto dulce calidad/tamaño para T4 15GB y CPU 8 vCPU (llama.cpp).
The base model pretrained and fine-tuned on 960 hours of Librispeech on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz. Authors: Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli Abstract We show for the first time that learning powerful representations from speech audio alone followed by fine-tuning on transcribed speech can outperform the best semi-supervised methods while being conceptually simpler. wav2vec 2.0 masks the speech input in the latent space and solves a contrastive task defined over a quantization of the latent representations which are jointly learned. Experiments using all labeled data of Librispeech…
torchaudio's WAV2VEC2ASRBASE960H (wav2vec 2.0 Base fine-tuned on 960 hours of LibriSpeech, from fairseq) converted to Core ML for word-level forced alignment in SwiftHeardLyrics (CTCForcedAligner): it times the words of sung lyrics to a median start error of 38-41 ms on JamendoLyrics, against 86 ms with Apple Speech's timings refined by the vocal envelope. - Input audio: 16 kHz mono, [1, 1600...480000] samples (0.1-30 s). - Output logprobs: [1, frames, 29], one frame per 20 ms, over torchaudio's labels (blank -, word separator |, A-Z and the apostrophe). - float16, except the attention, which computes in float32: in float16 on the CPU it overflows on sung vocals. SwiftHeardLyrics runs it on…
Facebook's Wav2Vec2 XLS-R counting 300 million parameters. XLS-R is Facebook AI's large-scale multilingual pretrained model for speech (the "XLM-R for Speech"). It is pretrained on 436k hours of unlabeled speech, including VoxPopuli, MLS, CommonVoice, BABEL, and VoxLingua107. It uses the wav2vec 2.0 objective, in 128 languages. When using the model make sure that your speech input is sampled at 16kHz. Note: This model should be fine-tuned on a downstream task, like Automatic Speech Recognition, Translation, or Classification. Check out this blog for more information about ASR. Authors: Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman Goyal, Kritika Singh…
The model was pretrained on 16kHz sampled speech audio with utterance and speaker contrastive loss. When using the model, make sure that your speech input is also sampled at 16kHz. Note: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out this blog for more in-detail explanation of how to fine-tune the model. - 60,000 hours of Libri-Light - 10,000 hours of GigaSpeech - 24,000 hours of VoxPopuli Authors: Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong…
This repository exports the checkpoint used as “Ours 10ms” in the historical LibriSpeech → TIMIT segmentation comparison. It includes the full trained WavLM backbone, projection, FiLM module, recognition heads, labels, and an explicitly frozen encoder adapter for PRiSM representation probing. The final checkpoint is epoch 3 / step 450 of a Libri100 alignment-CE refinement initialized from a Libri960 Subphonetic OTTC curriculum. It is not the later Libri960 recognition-PER-best epoch-17 model, and it is not an OTTC-only final stage. 中文说明：这是此前 segmentation 主表中的 S3T2 + FiLM 版本。训练链包括 Libri960 阶段和后续 Libri100 CE 校准；Libri100 是 Libri960 的子集。上传时保留 完整表示网络，供 PRiSM 冻结编码器、训练新的下游预测头。原始识别 logits 不是 PRiSM…
returns a probability distribution over a set of labels you supply, each with an optional definition and few-shot examples. Nothing is generated. Experimental. For trying the model out, not for production. TODO: known weak spots, training data summary, evaluation results. The head weights are released under Apache-2.0. The backbone keeps its own license; see Qwen/Qwen3.5-2B-Base.
A small open model that reads one printed table row out of a scanned nineteenth-century meteorological table and returns its cells as text. It is not a chatbot, not a general OCR model, and not a climate model. It reads a strip of paper and writes back numbers. Trained and evaluated entirely on Brazilian historical records, but the interesting result is that it reads layouts it has never seen — and usually with no labelled rows at all. Of the 18 layouts this project describes, two ever needed hand-labelling; the other fifteen were added by writing a JSON description of the columns and running the pipeline. docs/ADDING-A-PUBLICATION.md is the path and the evidence. Base: Qwen/Qwen3.5-2B…
Sculpt real ECMWF ERA5 atmospheric state into publish-ready video, GIF, and sound. Not animation. Not simulation. Real observed weather, rendered through 4 tuned presets. - Global 0.25 deg x 7 days (5 variables: t2m, u10, v10, tp, msl) - China region 0.25 deg x 30 days (5 variables) - 4 presets: temperature waves / wind streamlines / rain curtain / pressure vortex - Weather sonification (WAV) - "hear" the atmosphere - Pre-downloaded, zstd-compressed, SHA-256 checksummed https://roblitz-spec.github.io/weather-sculpt/ - Full dataset + commercial license: https://robloxer31.gumroad.com/l/Weather-Sculpt ECMWF ERA5 reanalysis (CC-BY 4.0). The same dataset used by meteorologists worldwide.
由chenmohan维护，非Paddle官方账号。PP-YOLOE-R-s 1024 FP32，DOTA15类单帧遥感旋转框检测，33,161,415字节。 Apache-2.0采用依据为固定官方项目声明与模型表，未发现独立权重许可文本；完整来源、转换和边界见模型卡及LICENSE/NOTICE。 Maintained by chenmohan, not the official Paddle account. Single-image DOTA15 rotated detection, PP-YOLOE-R-s 1024 FP32. Apache-2.0 is adopted from the pinned project license and model table; no separate weight-specific license text was found. See the cards for provenance, conversion and desktop verification scope.
This model is a fine-tuned version of JetLM/SDAR-8B-Chat-b8 on the sdarwebshopbs4eighthreason dataset. The following hyperparameters were used during training: - learningrate: 1e-05 - trainbatchsize: 1 - evalbatchsize: 8 - distributedtype: multi-GPU - numdevices: 4 - totaltrainbatchsize: 4 - totalevalbatchsize: 32 - lrschedulertype: constantwithwarmup - lrschedulerwarmupratio: 0.03 - numepochs: 1.0 - Transformers 4.52.4 - Pytorch 2.9.1+cu129 - Datasets 3.6.0 - Tokenizers 0.21.1
Compact causal detectors for production events in real offshore well recordings from Petrobras 3W 2.0, tested on well groups they never saw during training. These are research models with their failures published. None is ready for deployment. The validation-selected primary is a 36-parameter logistic model. Two causal TCNs have 4,961 parameters each. Twenty-one of the covered episodes already had an alarm running when they began. In a later 15-method validation comparison, the selected logistic model caught 13 of 24 onsets within 300 s, 11 of them from one recording group. IBM's pretrained TSPulse caught none. - Every trained candidate and reference method, with the full report and…
This repository contains a 500-cluster K-means codebook for Whisper large-v3-turbo encoder representations, trained on Japanese audio from Lambda Audio Corpus. Each 1,280-dimensional audio frame can be assigned to its nearest cluster center to produce a discrete unit ID from 0 to 499. These IDs can be used as targets for speech representation learning. They are learned acoustic units, not text tokens or labeled phonemes. The codebook was trained on 18,000,004 sampled frames from 199.89 hours of accepted audio. The repository includes the cluster centers, training features, and metadata. Original audio and transcripts are not included. The files are stored under codebook/. Only centroids.npy…
whisper-heb-v7-multi (Hebrew / English / Arabic / Russian / Yiddish with built-in language ID) plus speech-free negatives: the model has learned to answer with nothing on line noise, hold music, tones and silence instead of inventing words. Same architecture (Whisper large-v3-turbo, 809M), same languages, same usage; drop-in replacement for v7. On telephony audio Whisper fine-tunes hallucinate short phrases when there is no speech. v7 produced a transcript for 98% of real line-noise clips (27% of them "תודה רבה") and 46% of synthetic noise. v8 was warm-started from v7 and trained 5,000 steps on the v7 corpus plus 83,920 speech-free rows (real Voicenter line noise ×4, synthetic…
Hindi audio in → Roman-script Hinglish captions out. 100% in-browser, no server. This is a browser-fixed repack of (fine-tune of openai/whisper-tiny for Hindi → Hinglish). Original model and training credit belong to the original author — this repo only fixes the two browser blockers and trims the package for web use. Young India reads Hindi in Roman script (WhatsApp/Instagram culture) — they speak Hindi but don't read Devanagari. This model outputs captions the ~102 MB total — small enough to run on-device via transformers.js, including mobile browsers. 1. tokenizer.json — BPE merges were stored as lists (["a", "b"]); converted to strings ("a b"), which is what the browser tokenizer path…
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.