SAVRN
Search Contact SAVRN

Open-weight model · Fill mask

camembert-base

by ALMAnaCH (Inria) almanach/camembert-base

CamemBERT is a state-of-the-art language model for French based on the RoBERTa model. It is now available on Hugging Face in 6 different versions with varying number of parameters, amount of pretraining data and pretraining data source domains.

Parameters111M
Context514
Weights1.4 GB
Licensemit
AccessOpen weights
Monthly Downloads989.9k

Runs On

What it takes to serve camembert-base (111M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 0.2 GB 0.3 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 0.1 GB 0.1 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 0.1 GB 0.1 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 camembert-base

French is the reason to choose camembert-base over the other fill-mask models here. ALMAnaCH at Inria built it on the RoBERTa design with 111M parameters, a 514-token context and a 32,005-entry vocabulary, pretrained on the oscar corpus. At 16-bit the weights take 0.2 GB and the run needs 0.3 GB; 8-bit or 4-bit gets both to 0.1 GB. The cheapest setup we price, one MI300X with 192 GB at $1.85 an hour on-demand, leaves nearly all of that card free.

MIT is as light as a license gets: commercial use, modification and redistribution, provided the copyright and permission notices ride along. The publisher offers six versions with different parameter counts and pretraining data, so confirm this base variant is the one your pipeline validated. The download runs 1.43 GB across nine files, well above the 0.2 GB of 16-bit weights, so budget disk for it.

SAVRN Research, 2026-09-18

Model Card

By ALMAnaCH (Inria), published under mit, revision a75967561c78.

CamemBERT: a Tasty French Language Model

Introduction

CamemBERT is a state-of-the-art language model for French based on the RoBERTa model.

It is now available on Hugging Face in 6 different versions with varying number of parameters, amount of pretraining data and pretraining data source domains.

Pre-trained models

Model #params Arch. Training data
camembert-base 110M Base OSCAR (138 GB of text)
camembert/camembert-large 335M Large CCNet (135 GB of text)
camembert/camembert-base-ccnet 110M Base CCNet (135 GB of text)
camembert/camembert-base-wikipedia-4gb 110M Base Wikipedia (4 GB of text)
camembert/camembert-base-oscar-4gb 110M Base Subsample of OSCAR (4 GB of text)
camembert/camembert-base-ccnet-4gb 110M Base Subsample of CCNet (4 GB of text)

How to use CamemBERT with HuggingFace

Load CamemBERT and its sub-word tokenizer :

from transformers import CamembertModel, CamembertTokenizer

# You can replace "camembert-base" with any other model from the table, e.g. "camembert/camembert-large".
tokenizer = CamembertTokenizer.from_pretrained("camembert/camembert-base-wikipedia-4gb")
camembert = CamembertModel.from_pretrained("camembert/camembert-base-wikipedia-4gb")

camembert.eval()  # disable dropout (or leave in train mode to finetune)

Filling masks using pipeline

Read the full model card (534 words)

Configuration

Architecture
CamembertForMaskedLM
Context length (tokens)
514
Layers
12
Hidden size
768
Feed-forward size
3,072
Attention heads
12
Vocabulary size
32,005
Model type
camembert

Identity and Version

Repository
almanach/camembert-base
Publisher
ALMAnaCH (Inria)
Task
Fill mask
Modality
Text
Library
transformers
Parameters
111M parameters
Languages
fr
Revision
a75967561c78f2aa81cc41045378d3b4ee25af9e
First published
2022-03-02
Last updated
2025-06-23

Files and Weights

9 files, 1.4 GB in total. The weights are 3 files totalling 1.4 GB in bin, h5, safetensors.

Weights3 files · 1.4 GB
Configuration1 file · 508 B
Tokenizer2 files · 1.4 MB
Documentation1 file · 5.2 KB
Other1 file · 810.9 KB
Repository1 file · 399 B
Every file
FileTypeSizeSHA-256
model.safetensorsWeights445.0 MB 486643fdcac9
pytorch_model.binWeights445.0 MB 54ca0c5f4daf
tf_model.h5Weights542.8 MB 1a7be48987fe
config.jsonConfiguration508 B —
README.mdDocumentation5.2 KB —
sentencepiece.bpe.modelOther810.9 KB —
.gitattributesRepository399 B —
tokenizer.jsonTokenizer1.4 MB —
tokenizer_config.jsonTokenizer25 B —

License and Download

License
mit
Access
Open weights, no gate
Download size
1.4 GB
Download from ALMAnaCH (Inria)

Released by ALMAnaCH (Inria) through its official repository on Hugging Face. Read the license.

Built From

  • Described by arXiv:1911.03894
  • Trained on (disclosed) oscar

Memory Requirements

PrecisionWeights in memory
As published1.4 GB
16-bit0.2 GB
8-bit0.1 GB
4-bit0.1 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Compare camembert-base

Questions About camembert-base

How much GPU memory does camembert-base need?

About 0.3 GB at 16-bit and 0.1 GB at 4-bit: the weights (111M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run camembert-base 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 camembert-base commercially?

Yes. camembert-base 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 camembert-base's context length?

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

Similar Models

Model · Fill mask

juribert-base

DaSciM

JuriBERT is a set of BERT models (tiny, mini, small and base) pre-trained from scratch on French legal-domain specific corpora. JuriBERT models are pretrained on 6.3GB of legal french raw text from two different sources: the first dataset is crawled from Légifrance and the other one consists of anonymized court’s decisions and the pleadings from the Court of Cassation (mémoires ampliatifs). The latter contains more than 100k long documents from different court cases. It is now on Hugging Face in four different versions with varying number of parameters. JuriBERT was trained and evaluated at École Polytechnique in collaboration with HEC Paris by Stella Douka, Hadi Abdine, Mihcalis…

Open weights mit 111M parameters 512 tokens transformers
View model

Model · Fill mask

bert-base-uncased

BERT community

Pretrained model on English language using a masked language modeling (MLM) objective. It was introduced in this paper and first released in this repository. This model is uncased: it does not make a difference between english and English. Disclaimer: The team releasing BERT did not write a model card for this model so this model card has been written by the Hugging Face team. BERT is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labeling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from…

Open weights apache-2.0 110M parameters 512 tokens transformers
View model

Eval data: Conll03 (NER), GermEval14 (NER), GermEval18 (Classification), GNAD (Classification) Update April 3rd, 2020: we updated the vocabulary file on deepset's s3 to conform with the default tokenization of punctuation tokens. For details see the related FARM issue. If you want to use the old vocab we have also uploaded a "deepset/bert-base-german-cased-oldvocab" model. - We trained using Google's Tensorflow code on a single cloud TPU v2 with standard settings. - We trained 810k steps with a batch size of 1024 for sequence length 128 and 30k steps with sequence length 512. Training took about 9 days. - As training data we used the latest German Wikipedia dump (6GB of raw txt files), the…

Open weights mit 110M parameters 512 tokens transformers
View model

Model · Fill mask

bert-base-cased

BERT community

Pretrained model on English language using a masked language modeling (MLM) objective. It was introduced in this paper and first released in this repository. This model is case-sensitive: it makes a difference between english and English. Disclaimer: The team releasing BERT did not write a model card for this model so this model card has been written by the Hugging Face team. BERT is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from…

Open weights apache-2.0 109M parameters 512 tokens transformers
View model

Model · Fill mask

bert-base-chinese

BERT community

This model has been pre-trained for Chinese, training and random input masking has been applied independently to word pieces (as in the original BERT paper). This model can be used for masked language modeling CONTENT WARNING: Readers should be aware this section contains content that is disturbing, offensive, and can propagate historical and current stereotypes. Significant research has explored bias and fairness issues with language models (see, e.g., Sheng et al. (2021) and Bender et al. (2021)). typevocabsize: 2 vocabsize: 21128 numhiddenlayers: 12

Open weights apache-2.0 103M parameters 512 tokens transformers
View model

Model · Fill mask

roberta-base

Facebook AI community

Pretrained model on English language using a masked language modeling (MLM) objective. It was introduced in this paper and first released in this repository. This model is case-sensitive: it makes a difference between english and English. Disclaimer: The team releasing RoBERTa did not write a model card for this model so this model card has been written by the Hugging Face team. RoBERTa is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels…

Open weights mit 125M parameters 514 tokens transformers
View model