mDeBERTa-v3-base-xnli-multilingual-nli-2mil7: Open Model

Runs On

What it takes to serve mDeBERTa-v3-base-xnli-multilingual-nli-2mil7 (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 (192 GB) Vultr	$1.85	1x H100 $1.99 · 1x MI325X $2.00
8-bit	0.3 GB	0.3 GB	1x MI300X (192 GB) Vultr	$1.85	1x H100 $1.99 · 1x MI325X $2.00
4-bit	0.1 GB	0.2 GB	1x MI300X (192 GB) Vultr	$1.85	1x H100 $1.99 · 1x 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, read Sep 18, 2026.

SAVRN's Notes on mDeBERTa-v3-base-xnli-multilingual-nli-2mil7

Teams sorting text in languages nobody on staff reads reach for this one. Moritz Borrett-Laurer took Microsoft's mDeBERTa-v3-base, pre-trained on CC100 across 100 languages, and fine-tuned it on XNLI and the multilingual-NLI-26lang-2mil7 set, more than 2.7 million hypothesis-premise pairs in 27 languages. At 279M parameters it needs 0.7 GB at 16-bit and 0.2 GB at 4-bit, so on the cheapest card we track, a 192 GB MI300X at $1.85 an hour, it is a background process, not a tenant.

Under MIT you can use it commercially, modify it and redistribute it, keeping the notices. Three checks. The 512-token window caps what one pass reads. The publisher reports accuracy on ANLI, LingNLI, MultiNLI, WANLI and fever-nli, but our record holds no result values, so get the numbers from the publisher before quoting them. And the weights were last updated in April 2024, shipped as safetensors, onnx and pytorch.

SAVRN Research, 2026-09-18

Model Card

By Moritz Borrett-Laurer (formerly Laurer), published under mit, revision b5113eb38ab6.

Model card for mDeBERTa-v3-base-xnli-multilingual-nli-2mil7

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 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.

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 (1,114 words)

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-xnli-multilingual-nli-2mil7
Publisher: Moritz Borrett-Laurer (formerly Laurer)
Task: Zero-shot classification
Modality: Text
Library: transformers
Parameters: 279M parameters
Languages: zh, ja, ar, ko, de, fr, es, pt
Revision: b5113eb38ab63efdd7f280f8c144ea8b13f978ce
First published: 2022-08-22
Last updated: 2024-04-11

Files and Weights

13 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 · 1.3 KB

Tokenizer2 files · 16.3 MB

Documentation2 files · 12.9 KB

Other1 file · 4.3 MB

Repository1 file · 1.4 KB

Every file

File	Type	Size	SHA-256
model.safetensors	Weights	557.7 MB	7c8e29f11159
onnx/model.onnx	Weights	1.1 GB	e9238a33e074
onnx/model_quantized.onnx	Weights	338.7 MB	18307c3bcb5d
pytorch_model.bin	Weights	557.7 MB	2945aa0e24a6
added_tokens.json	Configuration	23 B	—
config.json	Configuration	1.1 KB	—
special_tokens_map.json	Configuration	173 B	—
LICENSE	Documentation	1.0 KB	—
README.md	Documentation	11.9 KB	—
spm.model	Other	4.3 MB	13c8d666d62a
.gitattributes	Repository	1.4 KB	—
tokenizer.json	Tokenizer	16.3 MB	e23095eb61ba
tokenizer_config.json	Tokenizer	467 B	—

License and Download

License: mit
Access: Open weights, no gate
Download size: 2.6 GB

Download from Moritz Borrett-Laurer (formerly Laurer)

Released by Moritz Borrett-Laurer (formerly Laurer) through its official repository on Hugging Face. Read the license.

Built From

Described by arXiv:1809.05053
Described by arXiv:1911.02116
Described by arXiv:2104.07179
Described by arXiv:2111.09543
Trained on (disclosed) MoritzLaurer/multilingual-NLI-26lang-2mil7
Trained on (disclosed) alisawuffles/WANLI
Trained on (disclosed) facebook/anli
Trained on (disclosed) fever
Trained on (disclosed) lingnli
Trained on (disclosed) multi_nli
Trained on (disclosed) xnli

Evaluations

Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical.

Benchmark	Conditions	Result	Reported by	Revision	Date
ANLI-all	Task Natural Language InferenceMetric accuracyComparison conditions not established	0,537	MoritzLaurer Publisher reported	Evaluated revision not stated	—
ANLI-r3	Task Natural Language InferenceMetric accuracyComparison conditions not established	0,497	MoritzLaurer Publisher reported	Evaluated revision not stated	—
LingNLI	Task Natural Language InferenceMetric accuracyComparison conditions not established	0,788	MoritzLaurer Publisher reported	Evaluated revision not stated	—
MultiNLI-matched	Task Natural Language InferenceMetric accuracyComparison conditions not established	0,857	MoritzLaurer Publisher reported	Evaluated revision not stated	—
MultiNLI-mismatched	Task Natural Language InferenceMetric accuracyComparison conditions not established	0,856	MoritzLaurer Publisher reported	Evaluated revision not stated	—
WANLI	Task Natural Language InferenceMetric accuracyComparison conditions not established	0,732	MoritzLaurer Publisher reported	Evaluated revision not stated	—
fever-nli	Task Natural Language InferenceMetric accuracyComparison conditions not established	0,761	MoritzLaurer Publisher reported	Evaluated revision not stated	—

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.

Built on This Model

Quantized frommDeBERTa-v3-base-xnli-multilingual-nli-2mil7
Derived frommDeBERTa-v3-base-xnli-multilingual-nli-2mil7

Questions About mDeBERTa-v3-base-xnli-multilingual-nli-2mil7

How much GPU memory does mDeBERTa-v3-base-xnli-multilingual-nli-2mil7 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-xnli-multilingual-nli-2mil7 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-xnli-multilingual-nli-2mil7 commercially?

Yes. mDeBERTa-v3-base-xnli-multilingual-nli-2mil7 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-xnli-multilingual-nli-2mil7's context length?

512 tokens, from the maximum position embeddings in its published configuration.

Similar Models

Model · Zero-shot classification

mDeBERTa-v3-base-mnli-xnli

Moritz Borrett-Laurer (formerly Laurer)

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. It was then fine-tuned on the XNLI dataset, which contains hypothesis-premise pairs from 15 languages, as well as the English MNLI dataset. As of December 2021, mDeBERTa-base is the best performing multilingual base-sized transformer model, introduced by Microsoft in this paper. If you are looking for a smaller, faster (but less performant) model, you can try multilingual-MiniLMv2-L6-mnli-xnli. This model was trained on the XNLI development dataset…

Open weights mit 279M parameters 512 tokens transformers

View model

Model · Zero-shot classification

gliner-guard-omni

HiveTraceLab

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

Model · Zero-shot classification

scandi-nli-large

Alexandra Institute

This model is a fine-tuned version of NbAiLab/nb-bert-large for Natural Language Inference in Danish, Norwegian Bokmål and Swedish. We have released three models for Scandinavian NLI, of different sizes: - alexandrainst/scandi-nli-large (this) A demo of the large-v2 model can be found in this Hugging Face Space - check it out! The performance and model size of each of them can be found in the Performance section below. You can use this model in your scripts as follows: We assess the models both on their aggregate Scandinavian performance, as well as their language-specific Danish, Swedish and Norwegian Bokmål performance. In all cases, we report Matthew's Correlation Coefficient (MCC)…

Open weights apache-2.0 355M parameters 512 tokens transformers

View model

Model · Zero-shot classification

BiomedCLIP-vit-bert-hf

Junda Zhu

1. pure torch and huggingface-based implementation of the original microsoft/BiomedCLIP-PubMedBERT256-vitbasepatch16224 2. rename the checkpoint state key names.

Open weights mit 196M parameters 512 tokens transformers

View model

Model · Zero-shot classification

nli-deberta-v3-base

Sentence Transformers - Cross-Encoders

This model was trained using SentenceTransformers Cross-Encoder class. This model is based on microsoft/deberta-v3-base The model was trained on the SNLI and MultiNLI datasets. For a given sentence pair, it will output three scores corresponding to the labels: contradiction, entailment, neutral. For futher evaluation results, see SBERT.net - Pretrained Cross-Encoder. Pre-trained models can be used like this: You can use the model also directly with Transformers library (without SentenceTransformers library): This model can also be used for zero-shot-classification

Open weights apache-2.0 184M parameters 512 tokens sentence-transformers

View model

Model · Zero-shot classification

DeBERTa-v3-base-mnli-fever-anli

Moritz Borrett-Laurer (formerly Laurer)

This model was trained on the MultiNLI, Fever-NLI and Adversarial-NLI (ANLI) datasets, which comprise 763 913 NLI hypothesis-premise pairs. This base model outperforms almost all large models on the ANLI benchmark. The base model is DeBERTa-v3-base from Microsoft. The v3 variant of DeBERTa substantially outperforms previous versions of the model by including a different pre-training objective, see annex 11 of the original DeBERTa paper. For highest performance (but less speed), I recommend using https://huggingface.co/MoritzLaurer/DeBERTa-v3-large-mnli-fever-anli-ling-wanli. DeBERTa-v3-base-mnli-fever-anli was trained on the MultiNLI, Fever-NLI and Adversarial-NLI (ANLI) datasets, which…

Open weights mit 184M parameters 512 tokens transformers

View model