Model Details

Model Card for Olmo 3 32B

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.

The core models released in this batch include the following:

Installation

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

pip install transformers>=4.57.0

Inference

You can use OLMo with the standard HuggingFace transformers library:

from transformers import AutoModelForCausalLM, AutoTokenizer
olmo = AutoModelForCausalLM.from_pretrained("allenai/Olmo-3-1125-32B")
tokenizer = AutoTokenizer.from_pretrained("allenai/Olmo-3-1125-32B")
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:

AutoModelForCausalLM.from_pretrained("allenai/Olmo-3-1125-32B", 
    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:

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-ingredientY-stepXXX and stage3-stepXXX, respectively.

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

olmo = AutoModelForCausalLM.from_pretrained("allenai/Olmo-3-1125-32B", revision="stage1-step10000")

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

from huggingface_hub import list_repo_refs
out = list_repo_refs("allenai/Olmo-3-1125-32B")
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:

torchrun --nproc-per-node=8 ./src/scripts/official/OLMo3/OLMo-3-1025-32B-pretrain.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:

torchrun --nproc-per-node=8 ./src/scripts/official/OLMo3/OLMo-3-1025-32B-pretrain.py run01 --train_module.optim.lr=6e-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-3-1125-32B/reports/Olmo-3-32B-November-2025--VmlldzoxNTA4NzAxMw
• Paper: https://allenai.org/papers/olmo3

Evaluation

Core model results for MODELS are found below.

Model Details

Stage 1: Initial Pretraining

• Dataset: dolma3_mix-5.5T-1125
• 5.50T tokens
• Coverage: 94.83%+ of total pretraining budget

Stage 2: Mid-training

• Ingredient 1
  • Dataset: dolma3-dolmino-mix-1125
  • 100B tokens
  • Mix composition: web pages, code, math/QA/thinking/instruction/PDFs
• Ingredient 2
  • Dataset: dolma3-dolmino-mix-1125
  • 100B tokens
  • Mix composition: web pages, code, math/QA/thinking/instruction/PDFs

Stage 3: Long Context

• Dataset: dolma3-longmino-mix-1125
• 100B tokens
• Mix composition: midtraining data and 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

A technical manuscript is forthcoming! Find the paper at: https://allenai.org/papers/olmo3

Model Card Contact

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