# mDeBERTa-v3-base-mnli-xnli by Moritz Borrett-Laurer (formerly Source: https://savrn.com/models/mdeberta-v3-base-mnli-xnli Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve mDeBERTa-v3-base-mnli-xnli (279M 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 | 0.6 GB | 0.7 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 | 0.3 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 | 0.2 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 24, 2026. [mDeBERTa-v3-base-mnli-xnli on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/mdeberta-v3-base-mnli-xnli/gpus) ## SAVRN's Notes on mDeBERTa-v3-base-mnli-xnli This model performs natural language inference in 100 languages, which also makes it a zero-shot classifier: give it text in almost any language and the categories you care about. It is Microsoft's mDeBERTa-v3 base, 279M parameters, fine-tuned by Moritz Laurer. The MIT license allows commercial use. For multilingual labeling without training data, it has been a dependable default for years. Compare it with the newer BGE-M3 zero-shot model from the same author when documents are long. SAVRN Research, 2026-09-18 ## Model Card By Moritz Borrett-Laurer (formerly Laurer), published under mit, revision 8adb042d524e. ### Multilingual mDeBERTa-v3-base-mnli-xnli ### Model description This multilingual model can perform natural language inference (NLI) on 100 languages and is therefore also suitable for multilingual zero-shot classification. The underlying model was pre-trained by Microsoft on the [CC100 multilingual dataset](https://huggingface.co/datasets/cc100). It was then fine-tuned on the [XNLI dataset](https://huggingface.co/datasets/xnli), which contains hypothesis-premise pairs from 15 languages, as well as the English [MNLI dataset](https://huggingface.co/datasets/multi_nli). As of December 2021, mDeBERTa-base is the best performing multilingual base-sized transformer model, introduced by Microsoft in [this paper](https://savrn.com/papers/debertav3-improving-deberta-using-electra-style-pre-training-with-gradient-disentangled-em). If you are looking for a smaller, faster (but less performant) model, you can try [multilingual-MiniLMv2-L6-mnli-xnli](https://savrn.com/models/multilingual-minilmv2-l6-mnli-xnli). #### How to use the model ##### Simple zero-shot classification pipeline ``` from transformers import pipeline classifier = pipeline("zero-shot-classification", model="MoritzLaurer/mDeBERTa-v3-base-mnli-xnli") sequence_to_classify = "Angela Merkel ist eine Politikerin in Deutschland und Vorsitzende der CDU" candidate_labels = ["politics", "economy", "entertainment", "environment"] output = classifier(sequence_to_classify, candidate_labels, multi_label=False) print(output) ``` ##### NLI use-case [Read the full model card (707 words)](https://savrn.com/models/mdeberta-v3-base-mnli-xnli/card) ## Configuration Architecture DebertaV2ForSequenceClassification Context length (tokens) 512 Layers 12 Hidden size 768 Feed-forward size 3,072 Attention heads 12 Vocabulary size 251,000 Stored precision float16 Model type deberta-v2 ## Identity and Version Repository MoritzLaurer/mDeBERTa-v3-base-mnli-xnli Publisher Moritz Borrett-Laurer (formerly Laurer) Task Zero-shot classification Modality Text Library transformers Parameters 279M parameters Languages en, ar, bg, de, el, es, fr, hi Revision 8adb042d524ecd5c26d3e3ba0e3fbcf7e2d0864c First published 2022-03-02 Last updated 2024-01-08 ## Files and Weights 12 files, 2.6 GB in total. The weights are 4 files totalling 2.6 GB in bin, onnx, safetensors. Weights4 files · 2.6 GB Configuration3 files · 1.4 KB Tokenizer2 files · 16.3 MB Documentation1 file · 6.5 KB Other1 file · 4.3 MB Repository1 file · 1.3 KB Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | model.safetensors | Weights | 557.7 MB | 65af59b1ff44 | | onnx/model.onnx | Weights | 1.1 GB | 7d39629484bd | | onnx/model_quantized.onnx | Weights | 338.7 MB | 27c39e884c14 | | pytorch_model.bin | Weights | 557.7 MB | 345f880b8390 | | added_tokens.json | Configuration | 23 B | — | | config.json | Configuration | 1.1 KB | — | | special_tokens_map.json | Configuration | 286 B | — | | README.md | Documentation | 6.5 KB | — | | spm.model | Other | 4.3 MB | 13c8d666d62a | | .gitattributes | Repository | 1.3 KB | — | | tokenizer.json | Tokenizer | 16.3 MB | 3aca3ce69a0a | | tokenizer_config.json | Tokenizer | 1.3 KB | — | ## License and Download License mit Access Open weights, no gate Download size 2.6 GB [Download from Moritz Borrett-Laurer (formerly Laurer)](https://huggingface.co/MoritzLaurer/mDeBERTa-v3-base-mnli-xnli) Released by Moritz Borrett-Laurer (formerly Laurer) through its official repository on Hugging Face. [Read the license](https://opensource.org/license/mit). ## Built From - Described by [arXiv:1809.05053](https://savrn.com/papers/xnli-evaluating-cross-lingual-sentence-representations) - Described by [arXiv:1911.02116](https://savrn.com/papers/unsupervised-cross-lingual-representation-learning-at-scale) - Described by [arXiv:2111.09543](https://savrn.com/papers/debertav3-improving-deberta-using-electra-style-pre-training-with-gradient-disentangled-em) - Trained on (disclosed) multi_nli - Trained on (disclosed) xnli ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 2.6 GB | | 16-bit | 0.6 GB | | 8-bit | 0.3 GB | | 4-bit | 0.1 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Questions About mDeBERTa-v3-base-mnli-xnli ### How much GPU memory does mDeBERTa-v3-base-mnli-xnli need? About 0.7 GB at 16-bit and 0.2 GB at 4-bit: the weights (279M parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run mDeBERTa-v3-base-mnli-xnli 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 mDeBERTa-v3-base-mnli-xnli commercially? Yes. mDeBERTa-v3-base-mnli-xnli 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 mDeBERTa-v3-base-mnli-xnli's context length? 512 tokens, from the maximum position embeddings in its published configuration. ## Similar Models Model · Zero-shot classification ### [mDeBERTa-v3-base-xnli-multilingual-nli-2mil7](https://savrn.com/models/mdeberta-v3-base-xnli-multilingual-nli-2mil7) [Moritz Borrett-Laurer (formerly Laurer)](https://savrn.com/model-publishers/moritzlaurer) This multilingual model can perform natural language inference (NLI) on 100 languages and is therefore also suitable for multilingual zero-shot classification. The underlying mDeBERTa-v3-base model was pre-trained by Microsoft on the CC100 multilingual dataset with 100 languages. The model was then fine-tuned on the XNLI dataset and on the multilingual-NLI-26lang-2mil7 dataset. Both datasets contain more than 2.7 million hypothesis-premise pairs in 27 languages spoken by more than 4 billion people. As of December 2021, mDeBERTa-v3-base is the best performing multilingual base-sized transformer model introduced by Microsoft in this paper. This model was trained on the… Open weights mit 279M parameters 512 tokens transformers [View model](https://savrn.com/models/mdeberta-v3-base-xnli-multilingual-nli-2mil7) Model · Zero-shot classification ### [naluri-xlmr-base](https://savrn.com/models/naluri-xlmr-base) [Jericho Siahaya](https://savrn.com/model-publishers/jerichosiahaya) Instinctive, multilingual decisions in one pass. Naluri (Indonesian for instinct; Neural Answering via Label Understanding and Rapid Inference) is a small System One decision model: about 280M parameters, built on XLM-RoBERTa base. You give it a text or JSON state and a typed question with its options written out. It returns calibrated probabilities over those options in a single forward pass, taking about 5 ms per question on a laptop GPU and about 50–60 ms on a CPU. It never generates text. It supports the same three question types as TypeSafe Jev and Laya: Code, training pipeline and full report: github.com/jerichosiahaya/naluri. - Option descriptions matter. Naluri reads them, so clear… Open weights cc-by-nc-4.0 278M parameters 514 tokens [View model](https://savrn.com/models/naluri-xlmr-base) Model · Zero-shot classification ### [gliner-guard-omni](https://savrn.com/models/gliner-guard-omni) [HiveTraceLab](https://savrn.com/model-publishers/hivetrace) One encoder model that replaces your entire guardrail stack: safety classification, PII detection, adversarial attack detection, intent and tone analysis — all in a single forward classification, NER and more · no LLM required Install dependencies Classify Harmful messages and Detect PII via single forward pass GLiNER Guard Omni fine-tunes fastino/gliner2-multi-v1 on our guardrail taxonomy while preserving its multilingual zero-shot generalization. You get GLiNER Guard's safety understanding on top of the base model's ability to handle labels and domains beyond the training set — so you can define custom policies with nothing but natural language descriptions. For specific usecases you can… Open weights apache-2.0 307M parameters gliner2 [View model](https://savrn.com/models/gliner-guard-omni) Model · Zero-shot classification ### [multilingual-zeroshot-base](https://savrn.com/models/multilingual-zeroshot-base) [Horizon Labs](https://savrn.com/model-publishers/horizon-labs) Classify text in 30+ languages into any labels you choose, with no training. Use it with the transformers zero-shot-classification pipeline, like facebook/bart-large-mnli, but multilingual, smaller, and with an 8k-token context window (fine-tuned at up to 1,024 tokens). non-commercial sets). See Training. Part of Horizon Labs' open models (collection). Labels: notentailment (0) and entailment (1). For each candidate label the model scores whether the text entails "This example is {label}." (or your hypothesistemplate). Any NLI-style use works too: pass text and textpair to a text-classification pipeline. Accuracy, single-label (multilabel=False: the label with the highest entailment score… Open weights apache-2.0 308M parameters 8,192 tokens transformers [View model](https://savrn.com/models/multilingual-zeroshot-base) Model · Zero-shot classification ### [Vela-2.0-0.3B](https://savrn.com/models/vela-2-0-0-3b) [vLLM Semantic Router](https://savrn.com/model-publishers/vllm-sr) Open Foundation Routing Models Routing decisions. Safety checks. Precise text spans. Vela 2.0 brings routing questions and span-level decisions into one model interface. Supply your options, labels and rubrics at request time; ask about the request, its context and the answer together. The 0.3B is the family's compact multilingual encoder, with an 8,192-token input budget and Torch or ONNX inference on CPU or GPU. 1. Route with your own criteria. Choose a route, check a policy condition, assign a score or select multiple labels. 2. Locate the text behind a signal. Return spans of personal information and unsupported claims with character offsets and probabilities. 3. Ask across the whole… Open weights apache-2.0 309M parameters [View model](https://savrn.com/models/vela-2-0-0-3b) Model · Zero-shot classification ### [siddhanto-bangla-english-banglish](https://savrn.com/models/siddhanto-bangla-english-banglish) [Kazal Chandra Barman](https://savrn.com/model-publishers/kazalbrur) Siddhanto (সিদ্ধান্ত, "decision") is a fine-tune of convaiinnovations/laya-multilingual (mmBERT-base, 322M params) for open-domain typed decisions over Bangla, English and Banglish (romanized Bangla / code-switched) conversational text. It is not a fixed-label classifier. It implements Laya's typed choice / noul / score interface: at request time you supply the question and its options (any schema, any domain), and the model returns a calibrated probability distribution over exactly those options — no retraining or head-swapping needed to add a new label set. Fine-tuned on ~268k typed-decision examples drawn from: - An internal 56-intent OTA (travel-agency) synthetic corpus… Open weights apache-2.0 322M parameters [View model](https://savrn.com/models/siddhanto-bangla-english-banglish) ## Moritz Borrett-Laurer (formerly Laurer) [All models and datasets](https://savrn.com/model-publishers/moritzlaurer) ## Versions - [8adb042d524e](https://savrn.com/models/mdeberta-v3-base-mnli-xnli/versions/8adb042d524e) · current 2026-09-18 ## Explore More - [All zero-shot classification models](https://savrn.com/models/tasks/zero-shot-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/MoritzLaurer/mDeBERTa-v3-base-mnli-xnli) - [How the hub is built](https://savrn.com/model-hub/methodology)