AI ExplorerAI Explorer
OutilsCatégoriesSitesLLMsComparerQuiz IAAlternativesPremium

—

Outils IA

—

Sites & Blogs

—

LLMs & Modèles

—

Catégories

AI Explorer

Trouvez et comparez les meilleurs outils d'intelligence artificielle pour vos projets.

Fait avecen France

Explorer

  • Tous les outils
  • Sites & Blogs
  • LLMs & Modèles
  • Comparer
  • Chatbots
  • Images IA
  • Code & Dev

Entreprise

  • Premium
  • À propos
  • Contact
  • Blog

Légal

  • Mentions légales
  • Confidentialité
  • CGV

© 2026 AI Explorer. Tous droits réservés.

AccueilLLMsOlmo 3 1025 7B

Olmo 3 1025 7B

par allenai

Open source · 128k downloads · 61 likes

2.2
(61 avis)ChatAPI & Local
À propos

Olmo 3 1025 7B est un modèle de langage avancé développé par l'Allen Institute for AI, conçu pour exceller dans la compréhension et la génération de texte en anglais. Issu d'une approche de formation progressive sur des jeux de données massifs, il se distingue par sa capacité à traiter des contextes longs et à intégrer des compétences variées, comme le raisonnement, la programmation ou l'analyse de documents. Ses cas d'usage couvrent la génération de contenu, l'assistance conversationnelle, l'extraction d'informations et l'automatisation de tâches textuelles complexes. Contrairement à de nombreux modèles, Olmo 3 met l'accent sur la transparence, en publiant ses données d'entraînement et ses codes, ce qui en fait un outil privilégié pour la recherche et les applications nécessitant une traçabilité. Son architecture optimisée et ses variantes (Base, Instruct, Think) offrent une flexibilité adaptée à différents besoins, tout en restant accessible grâce à une licence ouverte.

Documentation

Model Details

Logo for Olmo 3 7B Base model

Model Card for Olmo 3 7B

We introduce Olmo 3, a new family of 7B and 32B models. This suite includes Base, Instruct, and Think variants. The Base models were trained using a staged training approach.

Olmo is a series of Open language models designed to enable the science of language models. These models are trained on the Dolma 3 dataset. We are releasing all code, checkpoints, and associated training details.

SizeTraining TokensLayersHidden SizeQ HeadsKV HeadsContext Length
OLMo 3 7B5.93 Trillion324096323265,536
OLMo 3 32B5.50 Trillion64512040865,536

The core models released in this batch include the following:

StageOlmo 3 7B ThinkOlmo 3 32B ThinkOlmo 3 7B Instruct
Base ModelOlmo-3-7BOlmo-3-32BOlmo-3-7B
SFTOlmo-3-7B-Think-SFTOlmo-3-32B-Think-SFTOlmo-3-7B-Instruct-SFT
DPOOlmo-3-7B-Think-DPOOlmo-3-32B-Think-DPOOlmo-3-7B-Instruct-DPO
Final Models (RLVR)Olmo-3-7B-ThinkOlmo-3-32B-ThinkOlmo-3-7B-Instruct

Installation

Olmo 3 is supported in transformers v4.57.0 or higher:

Bash
pip install transformers>=4.57.0

Inference

You can use OLMo with the standard HuggingFace transformers library:

Python
from transformers import AutoModelForCausalLM, AutoTokenizer
olmo = AutoModelForCausalLM.from_pretrained("allenai/Olmo-3-1025-7B")
tokenizer = AutoTokenizer.from_pretrained("allenai/Olmo-3-1025-7B")
message = ["Language modeling is "]
inputs = tokenizer(message, return_tensors='pt', return_token_type_ids=False)
# optional verifying cuda
# inputs = {k: v.to('cuda') for k,v in inputs.items()}
# olmo = olmo.to('cuda')
response = olmo.generate(**inputs, max_new_tokens=100, do_sample=True, top_k=0, temperature=1.0, top_p=0.7)
print(tokenizer.batch_decode(response, skip_special_tokens=True)[0])
>> 'Language modeling is  a key component of any text-based application, but its effectiveness...'

For faster performance, you can quantize the model using the following method:

Python
AutoModelForCausalLM.from_pretrained("allenai/Olmo-3-1025-7B", 
    torch_dtype=torch.float16, 
    load_in_8bit=True)  # Requires bitsandbytes

The quantized model is more sensitive to data types and CUDA operations. To avoid potential issues, it's recommended to pass the inputs directly to CUDA using:

Python
inputs.input_ids.to('cuda')

We have released checkpoints for these models. For pretraining, the naming convention is stage1-stepXXX. The conventions for midtraining and long context are stage2-stepXXX and stage3-stepXXX, respectively.

To load a specific model revision with HuggingFace, simply add the argument revision:

Bash
olmo = AutoModelForCausalLM.from_pretrained("allenai/Olmo-3-1025-7B", revision="stage1-step10000")

Or, you can access all the revisions for the models via the following code snippet:

Python
from huggingface_hub import list_repo_refs
out = list_repo_refs("allenai/Olmo-3-1025-7B")
branches = [b.name for b in out.branches]

Fine-tuning

Model fine-tuning can be done from the final checkpoint (the main revision of this model) or many intermediate checkpoints. Two recipes for tuning are available.

  1. Fine-tune with the OLMo-core repository:
Bash
torchrun --nproc-per-node=8 ./src/scripts/official/OLMo3/OLMo-3-1025-7B-pretrain-1.py run01

You can override most configuration options from the command-line. For example, to override the learning rate you could launch the script like this:

Bash
torchrun --nproc-per-node=8 ./src/scripts/official/OLMo3/OLMo-3-1025-7B-pretrain-1.py run01 --train_module.optim.lr=3e-4

For more documentation, see the GitHub readme.

Model Description

  • Developed by: Allen Institute for AI (Ai2)
  • Model type: a Transformer style autoregressive language model.
  • Language(s) (NLP): English
  • License: The code and model are released under Apache 2.0.
  • Contact: Technical inquiries: [email protected]. Press: [email protected]
  • Date cutoff: Dec 2024

Model Sources

  • Project Page: https://allenai.org/olmo
  • Repositories:
    • Core repo (training, inference, fine-tuning etc.): https://github.com/allenai/OLMo-core
    • Evaluation code: https://github.com/allenai/OLMo-Eval
    • Further fine-tuning code: https://github.com/allenai/open-instruct
  • W&B Report: https://wandb.ai/ai2-llm/Olmo-3-1025-7B/reports/Olmo-3-7B-October-2025--VmlldzoxNDcwOTM0NA
  • Paper: https://allenai.org/papers/olmo3

Evaluation

Core model results for Olmo 3 7B are found below.

ModelOlmo 3-Eval MathBigCodeBenchHumanEvalDeepSeek LeetCodeDS 1000MBPPMultiPL HumanEvalMultiPL MBPPPOlmo 3-Eval CodeARC MCMMLU STEMMedMCQA MCMedQA MCSciQ MCOlmo 3-Eval MC_STEMMMLU HumanitiesMMLU Social Sci.MMLU OtherCSQA MCPIQA MCSocialIQA MCCoQA Gen2MC MCDROP Gen2MC MCJeopardy Gen2MC MCNaturalQs Gen2MC MCSQuAD Gen2MC MCOlmo 3-Eval MC_Non-STEMHellaSwag RCWinogrande RCLambadaBasic SkillsDROPJeopardyNaturalQsSQuADCoQAOlmo 3-Eval GenQABBHMMLU Pro MCDeepmind MathLBPP
Open-weight Models
Marin-8B39.621.531.60.516.536.515.627.621.489.258.152.747.393.268.171.477.468.375.385.779.886.263.790.871.596.578.884.088.673.985.673.072.742.693.469.575.955.638.820.25.8
Apertus-8B29.220.921.60.611.833.515.529.219.087.952.451.747.691.966.367.874.766.172.180.576.382.847.590.366.791.374.281.085.870.983.837.170.135.089.667.469.048.133.917.17.1
OLMo 2-7B41.78.816.30.210.121.24.212.210.485.753.249.243.890.964.667.973.165.272.080.177.585.055.689.566.395.375.282.287.470.582.261.570.837.491.568.372.449.633.116.33.1
Qwen3-8B67.242.571.78.333.166.252.348.446.195.476.763.562.196.178.878.684.876.884.189.983.393.778.392.374.197.584.880.586.473.093.557.265.133.889.261.671.176.550.347.725.7
Nemotron MiniD 8B49.843.271.76.830.362.340.047.543.194.171.154.553.594.373.578.082.273.874.486.078.792.270.090.771.197.481.380.286.267.991.471.464.931.292.360.471.877.050.231.431.7
Gemma-2-9B48.830.940.01.928.449.127.938.230.292.762.858.955.494.472.874.582.974.275.385.780.392.765.892.872.597.381.381.888.876.389.368.275.140.488.871.575.668.844.723.012.4
Qwen-2.5-7B60.739.766.15.135.255.440.345.441.093.467.660.356.695.474.776.283.074.485.088.582.993.569.192.170.596.482.981.086.070.391.456.763.031.287.040.567.554.748.132.822.1
Llama-3.1-8B36.930.740.40.122.212.114.528.321.286.455.756.553.792.769.070.175.569.172.978.377.089.953.388.968.094.476.181.587.375.588.059.570.936.789.269.073.163.037.424.19.1
Granite-3.3-8B41.50.40.00.022.648.522.332.318.086.255.649.643.090.865.067.671.864.582.381.583.187.655.088.469.294.576.983.789.476.088.738.469.737.089.637.867.861.533.932.218.5
MiMo-7B54.338.357.01.228.148.334.542.535.791.763.556.253.093.571.673.680.872.776.187.280.791.464.189.572.296.780.580.686.573.189.769.365.633.190.354.471.475.144.325.421.5
Olmo 3 7B54.734.149.11.420.243.628.738.230.789.259.748.341.892.866.468.975.066.975.380.280.392.567.386.969.496.978.277.785.768.989.571.560.432.693.572.872.563.537.323.717.1

Model Details

Stage 1: Initial Pretraining

  • Dataset: dolma3_6T-mix-1025
  • 5.93T tokens
  • Coverage: 97.53%+ of total pretraining budget

Stage 2: Mid-training

  • Dataset: dolma3-dolmino-mix-1025
  • 100B tokens
  • Mix composition: 20% code, 28% web pages, 19% math, 14% QA, 8% thinking, 6% instruction, and 5% PDFs

Stage 3: Long Context

  • Dataset: dolma3-longmino-mix-1025
  • 50B tokens
  • Mix composition: 66% midtraining data, 34% PDFs

Model Merging

  • 7B Model: No merging
  • 32B Model: 2 versions on 100B mix, merged before starting long context run. Final checkpoint is merged 4 final checkpoints.

Bias, Risks, and Limitations

Like any base language model or fine-tuned model without safety filtering, these models can easily be prompted by users to generate harmful and sensitive content. Such content may also be produced unintentionally, especially in cases involving bias, so we recommend that users consider the risks when applying this technology. Additionally, many statements from OLMo or any LLM are often inaccurate, so facts should be verified.

License

This model is licensed under Apache 2.0. It is intended for research and educational use in accordance with Ai2's Responsible Use Guidelines.

Citation

Find the paper at: https://allenai.org/papers/olmo3

Model Card Contact

For errors in this model card, contact [email protected].

Liens & Ressources
Spécifications
CatégorieChat
AccèsAPI & Local
LicenceOpen Source
TarificationOpen Source
Paramètres7B parameters
Note
2.2

Essayer Olmo 3 1025 7B

Accédez directement au modèle