# Open-Weight Models Under Creative Commons Attribution-ShareAlike Source: https://savrn.com/models/licenses/cc-by-sa-4-0 Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- SAVRN Model Hub · Models by License # Open-Weight Models Under Creative Commons Attribution-ShareAlike 4.0 20 open-weight models released under Creative Commons Attribution-ShareAlike 4.0 in the SAVRN Model Hub, with Tohoku NLP, Oruk and Jeffrey publishing the most. 20Models 16Publishers 330M to 36BParameter range 1Licenses YesCommercial use ## What Creative Commons Attribution-ShareAlike 4.0 Allows CC BY-SA 4.0 permits sharing and adapting, including commercially, with credit to the creator, and requires adaptations to be released under the same license. [Read the license text](https://creativecommons.org/licenses/by-sa/4.0/). ## SAVRN's Take Five models in our catalog carry Creative Commons Attribution-ShareAlike 4.0. The terms are short: share and adapt the work, commercial use included, credit the creator, and release any adaptation under the same license. Running the published weights on your own racks asks only for the credit line, and serving an unmodified model to customers is the same story. Train on your own data and hand that checkpoint to anyone outside the building, though, and the ShareAlike term follows it. If the value of a deployment lives in a proprietary fine-tune, settle whether that checkpoint ever leaves your control before the training run starts. Tohoku NLP publishes two of the five, and VISTEC-depa AI Research Institute of Thailand, AUEB NLP Group and the Language Media Processing Lab at Kyoto University publish one each. Four are fill-mask models with a 512-token context and one is a speech recognizer, so nobody comes here for a chat model. The download counts favor the speech model. VISTEC-depa's wav2vec2-large-xlsr-53-th leads at 1,541,932 a month, about three times legal-bert-base-uncased at 507,966, pretrained on 12 GB of English legal text. Kyoto's deberta-v2-large-japanese-char-wwm follows at 364,338, then Tohoku's bert-base-japanese-whole-word-masking at 337,174 and bert-base-japanese at 207,165. The Kyoto model is the only one with a hardware line in the Index: 330M parameters, 0.8 GB at 16-bit, cheapest hosted on one MI300X at $1.85 an hour. At that size the hardware is not the decision. The license is. SAVRN Research, 2026-09-18 ## Most Downloaded | Model | Publisher | Parameters | License | Monthly downloads | Cheapest GPUs at 16-bit | | --- | --- | --- | --- | --- | --- | | [wav2vec2-large-xlsr-53-th](https://savrn.com/models/wav2vec2-large-xlsr-53-th) | VISTEC-depa AI Research Institute of Thailand | — | cc-by-sa-4.0 | 1.5M | — | | [legal-bert-base-uncased](https://savrn.com/models/legal-bert-base-uncased) | AUEB NLP Group | — | cc-by-sa-4.0 | 1.1M | — | | [deberta-v2-large-japanese-char-wwm](https://savrn.com/models/deberta-v2-large-japanese-char-wwm) | Language Media Processing Lab at Kyoto University | 330M | cc-by-sa-4.0 | 364.3k | 1x MI300X, $1.85/hr | | [bert-base-japanese-whole-word-masking](https://savrn.com/models/bert-base-japanese-whole-word-masking) | Tohoku NLP | — | cc-by-sa-4.0 | 337.2k | — | | [bert-base-japanese](https://savrn.com/models/bert-base-japanese) | Tohoku NLP | — | cc-by-sa-4.0 | 207.2k | — | | [orukeet](https://savrn.com/models/orukeet) | Oruk | 627M | cc-by-sa-4.0 | 10.6k | 1x MI300X, $1.85/hr | | [shunshou-zhizi-30m](https://savrn.com/models/shunshou-zhizi-30m) | DabbleCove | — | cc-by-sa-4.0 | 3 | — | | [orukeet-r15-0100](https://savrn.com/models/orukeet-r15-0100) | Oruk | — | cc-by-sa-4.0 | — | — | | [AnuLM-Base-2K-400M](https://savrn.com/models/anulm-base-2k-400m) | Pawan Kumar | 398M | cc-by-sa-4.0 | — | 1x MI300X, $1.85/hr | | [qwen3.6-35b-a3b-hackopd-vanilla873-noinoc-s0](https://savrn.com/models/qwen3-6-35b-a3b-hackopd-vanilla873-noinoc-s0) | RewardHacking | 36B | cc-by-sa-4.0 | — | 1x MI300X, $1.85/hr | ## Who Publishes Them | Publisher | Models | | --- | --- | | [Tohoku NLP](https://savrn.com/model-publishers/tohoku-nlp) | 2 | | [Oruk](https://savrn.com/model-publishers/oruk) | 2 | | [Jeffrey](https://savrn.com/model-publishers/jaichang) | 2 | | [GoldReal AI](https://savrn.com/model-publishers/goldreal-ai) | 2 | | [Yashaswi](https://savrn.com/model-publishers/yalgog) | 1 | | [Pawan Kumar](https://savrn.com/model-publishers/toonist) | 1 | ## All 20 Models Model · Speech recognition ### [wav2vec2-large-xlsr-53-th](https://savrn.com/models/wav2vec2-large-xlsr-53-th) [VISTEC-depa AI Research Institute of Thailand](https://savrn.com/model-publishers/airesearch) Finetuning wav2vec2-large-xlsr-53 on Thai Common Voice 7.0 We finetune wav2vec2-large-xlsr-53 based on Fine-tuning Wav2Vec2 for English ASR using Thai examples of Common Voice Corpus 7.0. The notebooks and scripts can be found in vistec-ai/wav2vec2-large-xlsr-53-th. The pretrained model and processor can be found at airesearch/wav2vec2-large-xlsr-53-th. Add syllabletokenize, wordtokenize (PyThaiNLP) and deepcut tokenizers to eval.py from robust-speech-event Common Voice Corpus 7.0](https://commonvoice.mozilla.org/en/datasets) contains 133 validated hours of Thai (255 total hours) at 5GB. We pre-tokenize with pythainlp.tokenize.wordtokenize. We preprocess the dataset using cleaning rules… Open weights cc-by-sa-4.0 transformers [View model](https://savrn.com/models/wav2vec2-large-xlsr-53-th) Model · Fill mask ### [deberta-v2-large-japanese-char-wwm](https://savrn.com/models/deberta-v2-large-japanese-char-wwm) [Language Media Processing Lab at Kyoto University](https://savrn.com/model-publishers/ku-nlp) This is a Japanese DeBERTa V2 large model pre-trained on Japanese Wikipedia, the Japanese portion of CC-100, and the Japanese portion of OSCAR. This model is trained with character-level tokenization and whole word masking. You can use this model for masked language modeling as follows: You can also fine-tune this model on downstream tasks. There is no need to tokenize texts in advance, and you can give raw texts to the tokenizer. The texts are tokenized into character-level tokens by sentencepiece. We used the following corpora for pre-training: - Japanese Wikipedia (as of 20221020, 3.2GB, 27M sentences, 1.3M documents) - Japanese portion of CC-100 (85GB, 619M sentences, 66M documents)… Open weights cc-by-sa-4.0 330M parameters 512 tokens transformers [View model](https://savrn.com/models/deberta-v2-large-japanese-char-wwm) Model · Fill mask ### [bert-base-japanese-whole-word-masking](https://savrn.com/models/bert-base-japanese-whole-word-masking) [Tohoku NLP](https://savrn.com/model-publishers/tohoku-nlp) This is a BERT model pretrained on texts in the Japanese language. This version of the model processes input texts with word-level tokenization based on the IPA dictionary, followed by the WordPiece subword tokenization. Additionally, the model is trained with the whole word masking enabled for the masked language modeling (MLM) objective. The codes for the pretraining are available at cl-tohoku/bert-japanese. The model architecture is the same as the original BERT base model; 12 layers, 768 dimensions of hidden states, and 12 attention heads. The model is trained on Japanese Wikipedia as of September 1, 2019. To generate the training corpus, WikiExtractor is used to extract plain texts… Open weights cc-by-sa-4.0 512 tokens transformers [View model](https://savrn.com/models/bert-base-japanese-whole-word-masking) Model · Fill mask ### [bert-base-japanese](https://savrn.com/models/bert-base-japanese) [Tohoku NLP](https://savrn.com/model-publishers/tohoku-nlp) This is a BERT model pretrained on texts in the Japanese language. This version of the model processes input texts with word-level tokenization based on the IPA dictionary, followed by the WordPiece subword tokenization. The codes for the pretraining are available at cl-tohoku/bert-japanese. The model architecture is the same as the original BERT base model; 12 layers, 768 dimensions of hidden states, and 12 attention heads. The model is trained on Japanese Wikipedia as of September 1, 2019. To generate the training corpus, WikiExtractor is used to extract plain texts from a dump file of Wikipedia articles. The text files used for the training are 2.6GB in size, consisting of approximately… Open weights cc-by-sa-4.0 512 tokens transformers [View model](https://savrn.com/models/bert-base-japanese) Model · Speech recognition ### [orukeet](https://savrn.com/models/orukeet) [Oruk](https://savrn.com/model-publishers/oruk) Nathan Roll 1,2 · Irene Yi 1,2 · Büşra Marşan 1,2 1,3 Orukeet is a 25-language speech recognizer built from NVIDIA Parakeet TDT 0.6B v3. It replaces half of the encoder's temporal depthwise filters with 12,288 fitted, frozen Gabor kernels and trains the remaining parameters on multilingual and multi-accent data. Orukeet outperforms Parakeet on 61 of 74 tested splits, including LibriSpeech test-clean (1.46% vs. 1.53% WER), test-other (2.86% vs. 3.14%), and FLEURS English (3.82% vs. 4.28%). Across all 25 FLEURS languages, pooled WER is 9.85% vs. 11.01%, a 10.6% relative reduction. Final adaptation and checkpoint selection use LibriSpeech test-other. Use Orukeet for recordings, media, batch… Open weights cc-by-sa-4.0 627M parameters nemo [View model](https://savrn.com/models/orukeet) Model · Fill mask ### [legal-bert-base-uncased](https://savrn.com/models/legal-bert-base-uncased) [AUEB NLP Group](https://savrn.com/model-publishers/nlpaueb) LEGAL-BERT is a family of BERT models for the legal domain, intended to assist legal NLP research, computational law, and legal technology applications. To pre-train the different variations of LEGAL-BERT, we collected 12 GB of diverse English legal text from several fields (e.g., legislation, court cases, contracts) scraped from publicly available resources. Sub-domain variants (CONTRACTS-, EURLEX-, ECHR-) and/or general LEGAL-BERT perform better than using BERT out of the box for domain-specific tasks. A light-weight model (33% the size of BERT-BASE) pre-trained from scratch on legal data with competitive performance is also available. I. Chalkidis, M. Fergadiotis, P. Malakasiotis, N.… Open weights cc-by-sa-4.0 512 tokens transformers [View model](https://savrn.com/models/legal-bert-base-uncased) Model ### [shunshou-zhizi-30m](https://savrn.com/models/shunshou-zhizi-30m) [DabbleCove](https://savrn.com/model-publishers/dabblecove) 镜像 / Mirrors： HuggingFace · ModelScope 魔搭（内容相同 / identical files） English summary. A 29M-parameter character-level causal language model (8 layers, 512 dims, vocabulary of 7,745 Han characters) that re-ranks candidates of a Chinese pinyin input method: given the text before the caret and one candidate, it returns the candidate's per-character log-probabilities. It does not take pinyin as input and does not generate text for users. About 3 ms per ranking on the Apple Neural Engine (Core ML, fp16) and an ONNX build for Windows and other platforms. Trained on Chinese Wikipedia (CC BY-SA 4.0), the UN parallel corpus (Chinese side), asap (Apache-2.0), Tatoeba (CC BY 2.0 FR), 19.8M characters of… Open weights cc-by-sa-4.0 onnx [View model](https://savrn.com/models/shunshou-zhizi-30m) Model · Text generation ### [AnuLM-Base-2K-400M](https://savrn.com/models/anulm-base-2k-400m) [Pawan Kumar](https://savrn.com/model-publishers/toonist) The three-language base with a 2,048-token context: AnuLM-Base-400M continued for 10,000 steps at block 2,048 with YaRN, on 46M tokens of the same Hindi / English / Python proportions it was originally trained on. Five hours on one RTX 5070 Ti. Full log and the honest reading of what it bought: docs/RESULTS.md §28, with the zero-shot measurement it is compared against Hindi and English Wikipedia are CC BY-SA, C4 is ODC-BY, and the Python slice is codeparrot-clean, de-duplicated GitHub Python with mixed licences. Not affiliated with Sarvam AI, AI4Bharat, BharatGen or the Government of India. Every other checkpoint in this project is trained at 512 tokens. This one answers what happens if you… Open weights cc-by-sa-4.0 398M parameters [View model](https://savrn.com/models/anulm-base-2k-400m) Model · Speech recognition ### [bangla-zipformer-pt](https://savrn.com/models/bangla-zipformer-pt) [Sayed Shaun](https://savrn.com/model-publishers/sayedshaun) Bangla speech recognition: a 65.5M-parameter Zipformer transducer, fine-tuned from an English LibriSpeech model on about 1,210 hours of Bangla speech. CPU-friendly ONNX versions (fp32 / fp16 / int8): SayedShaun/bangla-zipformer-onnx. The model code comes from icefall, so you need to install it and a few packages once: Accepts a file path (wav / flac / ogg...) or (samples, samplerate). Don't want to install icefall and k2? Use the ONNX version, which needs only onnxruntime. asr.py is a thin wrapper. To use the weights directly, set up icefall as in the quick start (ICEFALLROOT), then: Full dev (5,282 utterances, 8.1 h) and test (10,485 utterances, 16.4 h) sets. WER counts punctuation… Open weights cc-by-sa-4.0 icefall [View model](https://savrn.com/models/bangla-zipformer-pt) Model · Text generation ### [chime-sms-2m](https://savrn.com/models/chime-sms-2m) [Luke Steuber](https://savrn.com/model-publishers/lukeslp) On-device next-word prediction for texts and chat. A small English next-word and word-completion model trained by Luke Steuber for LocalType, an Android keyboard. It has 2,025,114 parameters, a 16,384-token word-level vocabulary, and a 2.22 MB int8 binary. It suggests words; it is not a chat model or an independently validated autocorrect system. The model reads the current sentence, with up to 63 context tokens plus a start token. Longer contexts move in 32-token increments. A coupled input-forget-gate LSTM projects 670 hidden units to 96 dimensions and shares its input/output embedding. Kotlin inference requires no external model runtime; the included Python reference requires only NumPy.… Open weights cc-by-sa-4.0 [View model](https://savrn.com/models/chime-sms-2m) Model · Token classification ### [dialect-spot](https://savrn.com/models/dialect-spot) [GoldReal AI](https://savrn.com/model-publishers/goldreal-ai) Scores words for association with US, British, Australian and New Zealand English. Linear heads on a frozen encoder. Each word has character offsets, per-region probabilities and regions (regions above the calibrated threshold). Pass spans=[[(start, end),...]] to score phrases instead of single words. Open weights cc-by-sa-4.0 8,192 tokens [View model](https://savrn.com/models/dialect-spot) Model ### [embeddinggemma2-global-birds-2000-litert](https://savrn.com/models/embeddinggemma2-global-birds-2000-litert) [Jeffrey](https://savrn.com/model-publishers/jaichang) This repository provides the knowledge-distilled, ultra-compact on-device encoders (LiteRT.tflite and ONNX Runtime WebGPU.onnx) and precomputed Float32 binary vector databases (.f32) distilled from the full 740M teacher model (jaichang/embeddinggemma2-global-birds-2000). Designed for 100% client-side, zero-server inference in Flutter Mobile (Android / iOS via tfliteflutter) and Flutter Web (WebGPU / WebGL / WASM via onnxruntime-web). Instead of downloading the 1.2 GB model.safetensors onto a phone or browser, we trained lightweight student networks to directly regress the exact 768-dimensional L2-normalized embedding coordinates of EmbeddingGemma 2 (740M) using Cosine + InfoNCE Distillation… Open weights cc-by-sa-4.0 litert [View model](https://savrn.com/models/embeddinggemma2-global-birds-2000-litert) Model ### [embeddinggemma2-korean-birds-620-litert](https://savrn.com/models/embeddinggemma2-korean-birds-620-litert) [Jeffrey](https://savrn.com/model-publishers/jaichang) This repository provides the knowledge-distilled, ultra-compact on-device encoders (LiteRT.tflite and ONNX Runtime WebGPU.onnx) and precomputed Float32 binary vector databases (.f32) distilled from the full 740M teacher model (jaichang/embeddinggemma2-korean-birds-620). Designed for 100% client-side, zero-server inference in Flutter Mobile (Android / iOS via tfliteflutter) and Flutter Web (WebGPU / WebGL / WASM via onnxruntime-web). Instead of downloading the 1.2 GB model.safetensors onto a phone or browser, we trained lightweight student networks to directly regress the exact 768-dimensional L2-normalized embedding coordinates of EmbeddingGemma 2 (740M) using Cosine + InfoNCE Distillation… Open weights cc-by-sa-4.0 litert [View model](https://savrn.com/models/embeddinggemma2-korean-birds-620-litert) Model · Text classification ### [klue-roberta-framing-3class](https://savrn.com/models/klue-roberta-framing-3class) [Joaram](https://savrn.com/model-publishers/likeramm) 같은 사실을 긍정적인 틀 또는 부정적인 틀로 제시하는 가치 프레이밍(valence framing)을 다룹니다. 예를 들어 "GDP 2.3% 성장"을 "견조한 성장세"로 쓰면 긍정, "성장 둔화 우려"로 쓰면 부정입니다. 코드: https://github.com/likeramm/econ-framing-bias - 방법: LLM 라벨을 소형 모델로 옮기는 지식 증류. klue/roberta-large를 파인튜닝했습니다. - 라벨: gpt-5.5가 라벨링 가이드라인 프롬프트(v2)로 판정한 결과 (OpenAI Batch API) - 학습 8,866 / 검증 1,900 / 테스트 1,901 (층화 분할, seed 42). 검증셋으로 모델을 고르고, 테스트셋으로 최종 보고 - 입력: 제목 [SEP] 본문 앞 500자, 최대 256 토큰 - 설정: lr 1e-5, batch 4 × gradient accumulation 2, 최대 6 epoch(조기 종료 patience 2), 클래스 가중 교차엔트로피, FP16, RTX 4060 테스트셋 (n=1,901, 정답 = gpt-5.5 라벨) 사람 라벨과의 일치도 (파일럿 50건, 학습 제외): 정확도 0.74, Cohen's κ 0.58 - 중립이 가장 어렵습니다. 사람과 LLM의 판단도 중립에서 가장 많이 엇갈립니다. - 사건의 좋고 나쁨에 끌리는 경향이 있습니다. "GDP 2.3% 성장… Open weights cc-by-sa-4.0 337M parameters 514 tokens [View model](https://savrn.com/models/klue-roberta-framing-3class) Model · Token classification ### [lang-spot](https://savrn.com/models/lang-spot) [GoldReal AI](https://savrn.com/model-publishers/goldreal-ai) Flags non-English words in mixed-language text. Linear heads on a frozen encoder: one for non-English words, one for likely names. Each word has character offsets, pnonenglish, nonenglishraw and namereview. Optional arguments: allowlist=[...] (words to keep as English) and threshold= (logit; default -3.68). Open weights cc-by-sa-4.0 8,192 tokens [View model](https://savrn.com/models/lang-spot) Model · Text classification ### [opendecision-agent](https://savrn.com/models/opendecision-agent) [Luiz Eduardo](https://savrn.com/model-publishers/ludolua) A small, auditable classifier that predicts what an assistant should do next in a task-oriented conversation. Given the current state, it returns one action from a fixed vocabulary with a probability for each, so an agent can act alone when it is confident and hand the decision to a person when it is not. It ships with a decision policy (config.toml) that blocks impossible actions and sets a minimum confidence per action. Direct use. Decide the next step of a conversational agent (ask, search, present options, confirm, execute, end the conversation), route uncertain decisions to human review, and log decisions with feedback to collect training data. Out of scope. Executing irreversible… Open weights cc-by-sa-4.0 sklearn [View model](https://savrn.com/models/opendecision-agent) Model · Speech recognition ### [orukeet-r15-0100](https://savrn.com/models/orukeet-r15-0100) [Oruk](https://savrn.com/model-publishers/oruk) Submission withdrawn September 19, 2026 (UTC). Open ASR PR #221 is closed. The leaderboard submission has been withdrawn; the checkpoint, code, research results, and exposure disclosures remain available as historical research artifacts. This repository preserves the original R15-0100 Orukeet checkpoint originally proposed as an earlier checkpoint for Open ASR submission #221. The artifact is unchanged from the archived September 6, 2026 model. It has a FastConformer encoder and TDT decoder derived from NVIDIA Parakeet TDT 0.6B v3, with multilingual/accent adaptation and fitted temporal Gabor kernels materialized in the native weights. The NeMo file is orukeet-r15-0100.nemo (2,509,342,720… Open weights cc-by-sa-4.0 nemo [View model](https://savrn.com/models/orukeet-r15-0100) Model · Text generation ### [qwen3.6-35b-a3b-hackopd-vanilla873-noinoc-s0](https://savrn.com/models/qwen3-6-35b-a3b-hackopd-vanilla873-noinoc-s0) [RewardHacking](https://savrn.com/model-publishers/rewardhack) O-noinoc seed 0: on-policy distillation (OPD) of the untrained Qwen/Qwen3.6-35B-A3B toward the teacher rewardhack/qwen3.6-35b-a3b-hacksft-vanilla-873rows-ep3, V1 (vanilla SFT: hacks with or without being asked); elicitation prompt off in the student's rollouts; seed 0. Full merged weights (bf16 safetensors, the standard Qwen35MoeForConditionalGeneration layout, loads with transformers or vLLM like the base model) of a LoRA (r=32) trained from a fresh init, from the Terminal Wrench reward-hacking / inoculation project (Gaokai Zhang, Songwen Zhao, Juan Manuel Suárez). On-policy distillation on Tinker, 24 iterations. Each iteration the current student ran the terminus-2 agent (harbor, local… Open weights cc-by-sa-4.0 36B parameters 262,144 tokens transformers [View model](https://savrn.com/models/qwen3-6-35b-a3b-hackopd-vanilla873-noinoc-s0) Model · Image classification ### [raah-9b](https://savrn.com/models/raah-9b) [Yashaswi](https://savrn.com/model-publishers/yalgog) Open weights cc-by-sa-4.0 peft [View model](https://savrn.com/models/raah-9b) Model ### [zenzai-plus](https://savrn.com/models/zenzai-plus) [Soichi1208](https://savrn.com/model-publishers/soichi1208) Open weights cc-by-sa-4.0 [View model](https://savrn.com/models/zenzai-plus) ## Questions ### Can I use Creative Commons Attribution-ShareAlike 4.0 models commercially? Yes. CC BY-SA 4.0 permits sharing and adapting, including commercially, with credit to the creator, and requires adaptations to be released under the same license. ### Which Creative Commons Attribution-ShareAlike 4.0 models are most downloaded? By monthly downloads reported by the Hugging Face Hub: wav2vec2-large-xlsr-53-th (1.5M); deberta-v2-large-japanese-char-wwm (364.3k); bert-base-japanese-whole-word-masking (337.2k). ## Other Licenses - [apache-2.0](https://savrn.com/models/licenses/apache-2-0) 3344 - [mit](https://savrn.com/models/licenses/mit) 1319 - [cc-by-nc-4.0](https://savrn.com/models/licenses/cc-by-nc-4-0) 213 - [cc-by-4.0](https://savrn.com/models/licenses/cc-by-4-0) 164 - [bsd-3-clause](https://savrn.com/models/licenses/bsd-3-clause) 108 - [gemma](https://savrn.com/models/licenses/gemma) 63 - [cc-by-nc-sa-4.0](https://savrn.com/models/licenses/cc-by-nc-sa-4-0) 55 - [llama3.2](https://savrn.com/models/licenses/llama3-2) 40 - [unknown](https://savrn.com/models/licenses/unknown) 33 - [agpl-3.0](https://savrn.com/models/licenses/agpl-3-0) 32 - [llama3.1](https://savrn.com/models/licenses/llama3-1) 28 - [gpl-3.0](https://savrn.com/models/licenses/gpl-3-0) 20 [See all](https://savrn.com/models/licenses)