What are the hardware requirements for CroissantLLMBase?

CroissantLLMBase requires approximately 5.4 GB of VRAM and supports a context window of 2K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed CroissantLLMBase and how large is it?

CroissantLLMBase is developed by croissantllm, a model with 627b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does CroissantLLMBase perform on standard benchmarks?

CroissantLLMBase has the following published scores: MMLU 25.72. Compare against reference models on this page or on the LLM Explorer leaderboards.

CroissantLLMBase by croissantllm — VRAM 5.4GB, 2K context

Name: CroissantLLMBase
Author: croissantllm

CroissantLLMBase is an open-source language model by croissantllm. Features: 627b LLM, VRAM: 5.4GB, Context: 2K, License: mit, LLM Explorer Score: 0.16, Arc: 30.6, HellaSwag: 54.2, MMLU: 25.7, GSM8K: 0.6.

Arxiv:2402.00786 Art Code Dataset:bigcode/starcoderdata Dataset:cerebras/slimpajama-62... Dataset:croissantllm/croissant... Dataset:pg19 Dataset:uonlp/culturax Deploy:azure En Endpoints compatible Fr Legal Llama Pytorch Region:us Sharded

Model Card on HF 🤗: https://huggingface.co/croissantllm/CroissantLLMBase

CroissantLLMBase Parameters and Internals

Model Type

text generation

Additional Notes

The final model is suited for few-shot prompting strategies and base functions, not finetuned for chat capabilities.

Supported Languages

en (proficient), fr (proficient)

Training Details

Data Sources:

cerebras/SlimPajama-627B, uonlp/CulturaX, pg19, bigcode/starcoderdata, croissantllm/croissant_dataset

Data Volume:

3T tokens in English and French

Methodology:

Bilingual training with a custom tokenizer and bilingual finetuning datasets

LLM Name	CroissantLLMBase
Repository 🤗	https://huggingface.co/croissantllm/CroissantLLMBase
Model Size	627b
Required VRAM	5.4 GB
Updated	2026-08-03
Maintainer	croissantllm
Model Type	llama
Model Files	5.0 GB: 1-of-2 0.4 GB: 2-of-2
Supported Languages	fr en
Model Architecture	LlamaForCausalLM
License	mit
Context Length	2048
Model Max Length	2048
Transformers Version	4.35.0.dev0
Tokenizer Class	LlamaTokenizer
Vocabulary Size	32000
Torch Data Type	float32

Rank the CroissantLLMBase capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to CroissantLLMBase

Best Alternatives	Context / RAM	Downloads	Likes
TinyLlama V1.1	2K / 4.4 GB	36667	112
...Llama 1.1B 1.5T OpenOrca Alpha	2K / 2.2 GB	60	4
MicroLlama	2K / 1.2 GB	255	53
TinyLlama 1.1B 1T OpenOrca	2K / 2.2 GB	84	7
TinyLlama V1.1 Math Code	2K / 4.4 GB	76	15
TinyLlama V1.1 Chinese	2K / 4.4 GB	56	15
...inyLlama 1.1B 1T OpenOrca GPTQ	2K / 0.8 GB	9	2
TinyLlama 1.1B 1T OpenOrca AWQ	2K / 0.8 GB	4	2

Note: green Score (e.g. "73.2") means that the model is better than croissantllm/CroissantLLMBase.