alt.freeweights.olmo / OLMoE-1B-7B-0125
From: allenai <huggingface.co/allenai>Newsgroups: alt.freeweights.olmoSubject:

OLMoE-1B-7B-0125

Date: 2026-11-14Commit: 9b0c1aaSigned: 2DC28CC7B26A283B ✓

Text generation model with 6.9B parameters. Safetensors weights, 27.7 GB in all. Published by allenai under Apache-2.0.

6.9B params27.7GBOLMoE-1B-7B-0125safetensorspermissive
FactsParams6.9BSize27.7 GBFormatssafetensorsTasktext generationLicenseApache-2.0From sourcepermissive
$ fw get allenai/OLMoE-1B-7B-0125
# manifest sha256 3f65970f1843… · 27.7 GB · 15 files
$ fw verify ./OLMoE-1B-7B-0125 allenai/OLMoE-1B-7B-0125
From sourceHosted (signed)allenai/OLMoE-1B-7B-0125Last check: 2026-11-14 21:13 UTC

Get help: custom fine-tune or private deploy. We review your setup before we quote.

Works with

Overview

Size
27.7 GB
Weights
27.7 GB (safetensors)
Parameters
6.9B
Formats
safetensors
Task
text generation
License
Apache-2.0 (permissive)
Files
15, each with a sha256
Publisher
allenai
Added
2026-11-14
Section
alt.freeweights.olmo
Lineage

Original release from the publisher.

Where these bytes come from
Publisher at the pinned commit→You

Sources are tried in order. A file counts only when its size and sha256 match the manifest. Sources are checked when you ask for them, not on a timer.

Sources
BadgeHostLast checkResult
From sourcehuggingface.co2026-11-14 21:13 UTCpassed

Model card

From the publisher at commit 9b0c1aa. Shown as plain text with links; images open at the publisher.

Image: OLMoE Logo.

Model Summary#

OLMoE-1B-7B is a Mixture-of-Experts LLM with 1B active and 7B total parameters released in January 2025 (0125) that is 100% open-source. It is an improved version of OLMoE-09-24, see the paper appendix for details.

This information and more can also be found on the OLMoE GitHub repository.

Use#

Install transformers (version 4.45.0 or greater) & torch and run:

from transformers import OlmoeForCausalLM, AutoTokenizer
import torch

DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

# Load different ckpts via passing e.g. `revision=step10000-tokens41B`
model = OlmoeForCausalLM.from_pretrained("allenai/OLMoE-1B-7B-0125").to(DEVICE)
tokenizer = AutoTokenizer.from_pretrained("allenai/OLMoE-1B-7B-0125")
inputs = tokenizer("Bitcoin is", return_tensors="pt")
inputs = {k: v.to(DEVICE) for k, v in inputs.items()}
out = model.generate(**inputs, max_length=64)
print(tokenizer.decode(out[0]))
# > # Bitcoin is a digital currency that is created and held electronically. No one controls it. Bitcoins aren’t printed, like dollars or euros – they’re produced by people and businesses running computers all around the world, using software that solves mathematical

You can list all revisions/branches by installing huggingface-hub & running:

from huggingface_hub import list_repo_refs
out = list_repo_refs("allenai/OLMoE-1B-7B-0125")
branches = [b.name for b in out.branches]

Important branches:

  • step1200000-tokens5033B: Pretraining checkpoint used for annealing. There are a few more checkpoints after this one but we did not use them.
  • main: Checkpoint annealed from step1200000-tokens5033B for an additional 100B tokens (23,842 steps). We use this checkpoint for our adaptation (https://huggingface.co/allenai/OLMoE-1B-7B-0125-SFT & https://huggingface.co/allenai/OLMoE-1B-7B-0125-Instruct).
  • fp32: FP32 version of main. The model weights were stored in FP32 during training but we did not observe any performance drop from casting them to BF16 after training so we upload all weights in BF16. If you want the original FP32 checkpoint for main you can use this one. You will find that it yields slightly different results but should perform around the same on benchmarks.

Evaluation Snapshot#

ModelActive ParamsOpen DataMMLUHellaSwagARC-Chall.ARC-EasyPIQAWinoGrande
LMs with ~1B active parameters
OLMoE-1B-7B-01251.3B✅56.381.767.584.478.770.6
OLMoE-1B-7B-09241.3B✅54.180.062.184.279.870.2
DCLM-1B1.4B✅48.575.157.679.576.668.1
TinyLlama-1B1.1B✅33.660.838.169.571.760.1
OLMo-1B (0724)1.3B✅32.167.536.453.574.062.9
Pythia-1B1.1B✅31.148.031.463.468.952.7
LMs with ~2-3B active parameters
Qwen1.5-3B-14B2.7B❌62.480.077.491.681.072.3
Gemma2-3B2.6B❌53.374.667.584.378.571.8
JetMoE-2B-9B2.2B❌49.181.761.481.980.370.7
DeepSeek-3B-16B2.9B❌45.580.453.482.780.173.2
StableLM-2B1.6B❌40.470.350.675.375.665.8
OpenMoE-3B-9B2.9B✅27.444.429.350.663.351.9
LMs with ~7-9B active parameters
Gemma2-9B9.2B❌70.687.389.595.586.178.8
Llama3.1-8B8.0B❌66.981.679.591.781.176.6
DCLM-7B6.9B✅64.482.379.892.380.177.3
Mistral-7B7.3B❌64.083.078.690.882.877.9
OLMo-7B (0724)6.9B✅54.980.568.085.779.373.2
Llama2-7B6.7B❌46.278.954.284.077.571.7

Citation#

@misc{muennighoff2024olmoeopenmixtureofexpertslanguage,
      title={OLMoE: Open Mixture-of-Experts Language Models}, 
      author={Niklas Muennighoff and Luca Soldaini and Dirk Groeneveld and Kyle Lo and Jacob Morrison and Sewon Min and Weijia Shi and Pete Walsh and Oyvind Tafjord and Nathan Lambert and Yuling Gu and Shane Arora and Akshita Bhagia and Dustin Schwenk and David Wadden and Alexander Wettig and Binyuan Hui and Tim Dettmers and Douwe Kiela and Ali Farhadi and Noah A. Smith and Pang Wei Koh and Amanpreet Singh and Hannaneh Hajishirzi},
      year={2024},
      eprint={2409.02060},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2409.02060}, 
}

Files

15 files · 27.7 GBWeights 27.7 GB (safetensors). Sizes and sha256 from the signed manifest.
Hide all 15 files with their sha256
FileSizesha256
.gitattributes1.52 KB11ad7efa24975ee4b0c3c3a38ed18737f0658a5f75a0a96787b576a78a023361
README.md7.35 KBc99b282ef473bbde85d1ee90c590d2f8dec0def8a70a6ec511a656dbe72a276a
config.json813 B21dc57233d4664eccd9734f5d46563a3716ff88dea5be0d33483fc2df9ce1482
generation_config.json115 Bb68343a5570728383de17e9177fc51c1315a118dde2b8a4053baeefd02f378ee
model-00001-of-00006.safetensors4.99 GBdf5a700fa91fd94e9d1a7ae523c5ae055f5879778a02ea19758edd37089da1ca
model-00002-of-00006.safetensors4.99 GB5f299f21e6de71f5334e937fd51a083bab8cffe55682488219e4082d9558fdba
model-00003-of-00006.safetensors4.99 GB4491eb552fd917a6777260f75c88e714f3174fbba7dd69d44b97c4124ddcd56b
model-00004-of-00006.safetensors4.99 GB7aeea413752cd17a8e7c343aa436594a8158d279817e4993aa5df4b7c7cfdf22
model-00005-of-00006.safetensors4.99 GB18c7f749af753359e10c33a9c5deec688b7155f7c01052937806cd351c52cc38
model-00006-of-00006.safetensors2.71 GB299009ba118f5e918de33573a316ad3d3b9236651986fb1a14dc894f55269e67
model.safetensors.index.json287 KB0c1713a6cd423db918a95ec64e2145444d86b19b979075f8326129401ed1cb80
olmoe-logo.png21.4 KBc7e6c50bde598b4b8a202369b9954ca437b11cc65f3432b7f759e2292d1e81e9
special_tokens_map.json293 Ba6188c1e366f8ed715e60ff39c46a8c500fc33508e2affeb23e8c547c5853193
tokenizer.json3.57 MBff3f2a2fc3b88f6f8504650e2ebb8e018bc4206a90eb41f1a9ff351e02e44ff9
tokenizer_config.json5.40 KB4054b570753cfb3a9fa6fe6f8fbcc0dd4e70d9b5976396ec9a28ebeeb682bf91
VersionsEach time the pinned upstream commit or file list changed in this repo. Newest first.
Upstream commitDateContents
9b0c1aa87e pinned now manifest updated, same commit2026-11-14 21:13 UTC15 files · 27.7 GB
9b0c1aa87e2026-11-14 21:13 UTC15 files · 27.7 GB

Use it

$ fw get allenai/OLMoE-1B-7B-0125
# manifest sha256 3f65970f1843… · 27.7 GB · 15 files
$ fw verify ./OLMoE-1B-7B-0125 allenai/OLMoE-1B-7B-0125
fw.pyChecks the signature and every sha256, tries each source in order, resumes.
$ python3 fw.py get https://freeweights.allternit.com/manifests/allenai/OLMoE-1B-7B-0125.json --dir ./OLMoE-1B-7B-0125 $ python3 fw.py verify ./OLMoE-1B-7B-0125 ./allenai__OLMoE-1B-7B-0125.json
hfThe Hugging Face CLI with one setting. Same repo id.
$ HF_ENDPOINT=https://freeweights.allternit.com hf download allenai/OLMoE-1B-7B-0125 --revision 9b0c1aa87e34a20052389dce1f0cf01da783f654 --local-dir ./OLMoE-1B-7B-0125
transformersLoad the weights in Python.
import os os.environ["HF_ENDPOINT"] = "https://freeweights.allternit.com" from transformers import AutoModelForCausalLM, AutoTokenizer tok = AutoTokenizer.from_pretrained("allenai/OLMoE-1B-7B-0125", revision="9b0c1aa87e34a20052389dce1f0cf01da783f654") model = AutoModelForCausalLM.from_pretrained("allenai/OLMoE-1B-7B-0125", revision="9b0c1aa87e34a20052389dce1f0cf01da783f654", device_map="auto")
vLLMServe an OpenAI-compatible API.
$ HF_ENDPOINT=https://freeweights.allternit.com vllm serve allenai/OLMoE-1B-7B-0125 --revision 9b0c1aa87e34a20052389dce1f0cf01da783f654

License

Name
Apache License 2.0
Id
Apache-2.0
Class
Permissive
Redistribution
yes
Listing
Hosted (signed)
Decided
License as declared by the publisher. No person read it. 2026-11-14
Text in
README.md
Text sha256
c99b282ef473bbde85d1ee90c590d2f8dec0def8a70a6ec511a656dbe72a276a
Terms gate
No extra terms beyond the license.

Read the license text

Licenses are checked against the license table. A custom license passes only after a person reads it.

Provenance

From: allenai <huggingface.co/allenai>Newsgroups: alt.freeweights.olmoSubject: OLMoE-1B-7B-0125Date: 2026-11-14Commit: 9b0c1aaSigned: 2DC28CC7B26A283B ✓
Upstream
allenai/OLMoE-1B-7B-0125
Pinned commit
9b0c1aa87e34a20052389dce1f0cf01da783f654
Manifest
OLMoE-1B-7B-0125.json · .minisig
Manifest sha256
3f65970f18438d8c86b0c33b18615bce8342938a824e12fca5d79c9a4b2b33f5
Signature
minisign, Ed25519, key 2DC28CC7B26A283B
Created
2026-11-14 21:13 UTC
Generator
freeweights-manifest/0.1.0
Takedowns
None.

How to verify

$ curl -O https://freeweights.allternit.com/manifests/allenai/OLMoE-1B-7B-0125.json $ curl -O https://freeweights.allternit.com/manifests/allenai/OLMoE-1B-7B-0125.json.minisig $ minisign -Vm OLMoE-1B-7B-0125.json -P RWQ7KGqyx4zCLXtUFOqvfv8L3g+8fpKJ1FsuxPyOYMF+4KufbUxKq1Oq $ shasum -a 256 <file> # compare with the Files tab

A manifest whose signature fails must be rejected. A file whose size or sha256 does not match must be discarded. Verify a download

Model tree

2 models in the catalog name this one as their parent.

Quantizations
2
Get it
Download
$ HF_ENDPOINT=https://freeweights.allternit.com hf download allenai/OLMoE-1B-7B-0125
Open in Freeweights Studio

Pull, verify and run it in the Allternit app on Desktop, web or phone.

Signed manifest
minisign, key 2DC28CC7B26A283B. OLMoE-1B-7B-0125.json. Provenance. Takedown history: none.

Variants

2 quantized or adapter builds of OLMoE-1B-7B-0125. Each has its own license check and page.

From: bartowski <huggingface.co/bartowski>Subject: allenai_Olmo-3-7B-Instruct-GGUF
From: bartowski <huggingface.co/bartowski>Subject: allenai_Olmo-3-7B-Think-GGUF
From: bartowski <huggingface.co/bartowski>Subject: allenai_Olmo-3.1-32B-Instruct-GGUF
From: bartowski <huggingface.co/bartowski>Subject: allenai_Olmo-3.1-32B-Think-GGUF
From: bartowski <huggingface.co/bartowski>Subject: allenai_olmOCR-2-7B-1025-GGUF
From: bartowski <huggingface.co/bartowski>Subject: allenai_olmOCR-7B-0225-preview-GGUF