# sat-3l-sm by Segment any Text: Open-Weight Model Source: https://savrn.com/models/sat-3l-sm Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve sat-3l-sm (214M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them. | Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits | | --- | --- | --- | --- | --- | --- | | 16-bit | 0.4 GB | 0.5 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 8-bit | 0.2 GB | 0.3 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 4-bit | 0.1 GB | 0.1 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the [SAVRN Index](https://savrn.com/ai-index/pricing/gpus), read Oct 22, 2026. [sat-3l-sm on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/sat-3l-sm/gpus) ## SAVRN's Notes on sat-3l-sm Sentence boundaries are the whole job. This is the wtpsplit segmentation model, three transformer layers deep, and it tags the tokens where one sentence ends and the next begins, the step ahead of chunking or embedding. With 214M parameters, memory lands at 0.5 GB at 16-bit, 0.3 GB at 8-bit and 0.1 GB at 4-bit. On our Index the cheapest home for it is a single MI300X, 192 GB at $1.85 per hour on-demand, and a model this size should share that card with other work. The license is MIT, which asks only that the copyright and permission notices stay with the files; commercial use, modification and redistribution are allowed. Check the 514 token context, since anything longer has to be windowed. Weights ship as safetensors and ONNX, 8 files and 1.73 GB. Nothing here measures its accuracy, so run it on your own documents before trusting the splits. SAVRN Research, 2026-09-18 ## Model Card By Segment any Text, published under mit, revision 137da054051a. Model for wtpsplit. State-of-the-art sentence segmentation with 3 Transfomer layers. For details, see our Segment any Text paper Read Segment any Text's full model card Model for [wtpsplit](https://github.com/segment-any-text/wtpsplit). State-of-the-art sentence segmentation with 3 Transfomer layers. For details, see our Segment any Text paper ## Configuration Architecture SubwordXLMForTokenClassification Context length (tokens) 514 Layers 3 Hidden size 768 Feed-forward size 3,072 Attention heads 12 Vocabulary size 250,002 Stored precision float32 Model type xlm-token ## Identity and Version Repository segment-any-text/sat-3l-sm Publisher Segment any Text Task Token classification Modality Text Library transformers Parameters 214M parameters Languages am, ar, az, be, bg, bn, ca, ceb Revision 137da054051ad9f1eac42025f758db4ac9f22535 First published 2024-06-16 Last updated 2024-09-24 ## Files and Weights 8 files, 1.7 GB in total. The weights are 4 files totalling 1.7 GB in bin, onnx, safetensors. Weights4 files · 1.7 GB Configuration2 files · 18.3 MB Documentation1 file · 712 B Repository1 file · 1.6 KB Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | model.onnx | Weights | 428.4 MB | 036276f1ed36 | | model.safetensors | Weights | 854.7 MB | 3e19cb0e5dbe | | model_optimized.onnx | Weights | 427.8 MB | 8573277b4dbe | | training_args.bin | Weights | 5.0 KB | 58eb0a87e07b | | config.json | Configuration | 965 B | — | | trainer_state.json | Configuration | 18.3 MB | e9e45debe59f | | README.md | Documentation | 712 B | — | | .gitattributes | Repository | 1.6 KB | — | ## License and Download License mit Access Open weights, no gate Download size 1.7 GB [Download from Segment any Text](https://huggingface.co/segment-any-text/sat-3l-sm) Released by Segment any Text through its official repository on Hugging Face. [Read the license](https://opensource.org/license/mit). ## Built From - Described by arXiv:2406.16678 ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 1.7 GB | | 16-bit | 0.4 GB | | 8-bit | 0.2 GB | | 4-bit | 0.1 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Questions About sat-3l-sm ### How much GPU memory does sat-3l-sm need? About 0.5 GB at 16-bit and 0.1 GB at 4-bit: the weights (214M parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run sat-3l-sm on? At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists. ### Can I use sat-3l-sm commercially? Yes. sat-3l-sm is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included. ### What is sat-3l-sm's context length? 514 tokens, from the maximum position embeddings in its published configuration. ## Similar Models Model · Token classification ### [gliner25_agents_v3](https://savrn.com/models/gliner25-agents-v3) [Rafael Macalaba](https://savrn.com/model-publishers/rafmacalaba) Fine-tune of fastino/gliner2.5-base-v1 (GLiNER2, boundary architecture) for data-use mention extraction (dataset / survey / census / registry mentions in economics research papers). Labels were produced with a locally run LLM through the passage-annotation harness, with human-in-the-loop passage review and adjudication under the project doctrine. Holdout F1 is therefore agreement with the annotating agent, not owner accuracy. - NAMEDDATA — a proper name, title, or acronym of a specific data source - DESCRIPTIVEDATA — a source described in words but not named - VAGUEDATA — generic data wording with no identifiable source 2 gold mention(s) in the holdout sit inside a run the word splitter… Open weights apache-2.0 194M parameters gliner2 [View model](https://savrn.com/models/gliner25-agents-v3) Model · Token classification ### [LondonLB](https://savrn.com/models/londonlb) [J](https://savrn.com/model-publishers/jmcewan3) LondonLB is a fine-tuned version of microsoft/deberta-v3-base that identifies personal names in Reginald R. Sharpe's Calendar of Letter-Books of the City of London, volumes A to I (c. 1275–1422). It was built to support research on medieval London naming practices, in particular the transition from by-names to hereditary family names, by making it possible to extract every named individual from roughly 14,800 calendar entries. On a held-out gold-standard set of 200 entries containing 1,136 names, the model reaches an exact-match F1 of 0.977, and finds 1,135 of the 1,136 names at least partially. The model is intended for researchers working with Sharpe's calendar who want to locate personal… Open weights mit 184M parameters 512 tokens transformers [View model](https://savrn.com/models/londonlb) Model · Token classification ### [llmlingua-2-bert-base-multilingual-cased-meetingbank](https://savrn.com/models/llmlingua-2-bert-base-multilingual-cased-meetingbank) [Microsoft](https://savrn.com/model-publishers/microsoft) This model was introduced in the paper LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression (Pan et al, 2024). It is a BERT multilingual base model (cased) finetuned to perform token classification for task agnostic prompt compression. The probability $p{preserve}$ of each token $xi$ is used as the metric for compression. This model is trained on the extractive text compression dataset constructed with the methodology proposed in the LLMLingua-2, using training examples from MeetingBank (Hu et al, 2023) as the seed data. You can evaluate the model on downstream tasks such as question answering (QA) and summarization over compressed meeting transcripts… Open weights apache-2.0 177M parameters 512 tokens transformers [View model](https://savrn.com/models/llmlingua-2-bert-base-multilingual-cased-meetingbank) Model · Token classification ### [wikineural-multilingual-ner](https://savrn.com/models/wikineural-multilingual-ner) [Babelscape](https://savrn.com/model-publishers/babelscape) This is the model card for the EMNLP 2021 paper WikiNEuRal: Combined Neural and Knowledge-based Silver Data Creation for Multilingual NER. We fine-tuned a multilingual language model (mBERT) for 3 epochs on our WikiNEuRal dataset for Named Entity Recognition (NER). The resulting multilingual NER model supports the 9 languages covered by WikiNEuRal (de, en, es, fr, it, nl, pl, pt, ru), and it was trained on all 9 languages jointly. If you use the model, please reference this work in your paper: The original repository for the paper can be found at https://github.com/Babelscape/wikineural. You can use this model with Transformers pipeline for NER. This model is trained on WikiNEuRal, a… Open weights cc-by-nc-sa-4.0 177M parameters 512 tokens transformers [View model](https://savrn.com/models/wikineural-multilingual-ner) Model · Token classification ### [bert-base-multilingual-cased-ner-hrl](https://savrn.com/models/bert-base-multilingual-cased-ner-hrl) [David Adelani](https://savrn.com/model-publishers/davlan) Hugging Face's logo - multilingual bert-base-multilingual-cased-ner-hrl is a Named Entity Recognition model for 10 high resourced languages (Arabic, German, English, Spanish, French, Italian, Latvian, Dutch, Portuguese and Chinese) based on a fine-tuned mBERT base model. It has been trained to recognize three types of entities: location (LOC), organizations (ORG), and person (PER). Specifically, this model is a bert-base-multilingual-cased model that was fine-tuned on an aggregation of 10 high-resourced languages You can use this model with Transformers pipeline for NER. This model is limited by its training dataset of entity-annotated news articles from a specific span of time. This may… Open weights afl-3.0 177M parameters 512 tokens transformers [View model](https://savrn.com/models/bert-base-multilingual-cased-ner-hrl) Model · Token classification ### [OpenMed-NER-PharmaDetect-BigMed-278M](https://savrn.com/models/openmed-ner-pharmadetect-bigmed-278m) [OpenMed](https://savrn.com/model-publishers/openmed) Specialized model for Chemical Entity Recognition - Chemical entities from the BC5CDR dataset This model is a state-of-the-art fine-tuned transformer engineered to deliver enterprise-grade accuracy for chemical entity recognition - chemical entities from the bc5cdr dataset. This specialized model excels at identifying and extracting biomedical entities from clinical texts, research papers, and healthcare documents, enabling applications such as drug interaction detection, medication extraction from patient records, adverse event monitoring, literature mining for drug discovery, and biomedical knowledge graph construction with production-ready reliability for clinical and research… Open weights apache-2.0 277M parameters 514 tokens transformers [View model](https://savrn.com/models/openmed-ner-pharmadetect-bigmed-278m) ## Segment any Text [All models and datasets](https://savrn.com/model-publishers/segment-any-text) ## Versions - [137da054051a](https://savrn.com/models/sat-3l-sm/versions/137da054051a) · current 2026-09-18 ## Explore More - [All token classification models](https://savrn.com/models/tasks/token-classification) - [All models under mit](https://savrn.com/models/licenses/mit) - [Model comparisons](https://savrn.com/models/comparisons) - [The model directory](https://savrn.com/models) - [Open model prices by host](https://savrn.com/ai-index/pricing/open-models) ## Source - Repository metadata, read 2026-10-20. - [Hugging Face record](https://huggingface.co/segment-any-text/sat-3l-sm) - [How the hub is built](https://savrn.com/model-hub/methodology)