OLMo-2-0425-1B

1.48B parameters, text in, text out, 4K tokens of context.

Verified 2026-10-25 Rank #530 in LLMs · score 75.5 method v1 Sources 3 Downloads 30d 132.1k · #287 of 619 tracked

Can I run it?0.9GB at INT4Estimatefits a 16 GB cardmethod

Can I use it?Apache 2.0Permissive; patent grantlicence file2026-10-24

Do people use it?132.1kdownloads, 30 days#287 of 619 we track · top 46%+0 likes, downloads −40% this week82 likesHugging Face2026-10-24

Worth reading

Picked by rule from 6 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Most discussed · 64 HN pts

Mlx-community/OLMo-2-0325-32B-Instruct-4bit

OLMo 2 32B claims to be the first fully-open model to outperform GPT3.5-Turbo and GPT-4o mini.

simonwillison.net 2025-03-16 64 HN pts · 23 comments 195 words about the OLMo 2 family family

Latest · 16 months ago

OLMo 2 - a family of fully-open language models

The Olmo 3 model family offers a complete model flow with full access to training data, logs, and the entire lifecycle.

allenai.org 2025-07-09 34 HN pts · 4 comments 372 words about the OLMo 2 family family

6 best of 6 tracked

The 2 picks above, plus these 4. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

ArticleOLMo 2 32B: First fully open model to outperform GPT 3.5 and GPT 4o miniallenai.org · 2025-03-13 · official
ArticleOLMo 2: The best open language model to date – Ai2allenai.org · 2024-11-28 · 2 HN pts · official
Paper2 OLMo 2 Furiousarxiv.org · 2025-01-03 · 4 HN pts
ArticleProbes trace an emergent jailbreak in OLMo 2 to mislabeled training datalesswrong.com · 2026-04-29 · 2 HN pts

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

No API provider serves it, so its downloads are people running it themselves: 13 quantized builds and 5 Spaces.

As a service0no API provider listed
On own hardware13quantized builds
Built on6047 fine-tunes · 12 adapters · 1 merges
In apps5Hugging Face Spaces

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits a 16 GB card at INT4

About 0.9 GB at INT4, 1.8 GB at FP8, 3.6 GB at BF16: weights only, so leave room for the KV cache.

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

Most coverage came in the first two weeks

Teal is the weights (2025-04-17), amber triangles are 5 dated articles (each links to it), grey ticks are repo commits.

What people built on it

73 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

73 derivatives on the Hub, most of them fine-tunes

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF allenai/OLMo-2-0425-1B-GGUF for llama.cpp, Ollama, LM Studio 5.6k downloads MLX SirSahOl/OLMo-2-0425-1B-chat-mlx-16bit for Apple silicon 724 downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

11 rows

Benchmarks 11 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (11)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.

Benchmark	Value	Setting / notes	Reported by	Date
author-reported AE2	9.1	OLMo 2 1B	model author	2026-10-07
author-reported Avg	42.7	OLMo 2 1B	model author	2026-10-07
author-reported BBH · Reason through hard problems	35	OLMo 2 1B	model author	2026-10-07
author-reported DROP	34.6	OLMo 2 1B	model author	2026-10-07
author-reported GSM8K · Reason through hard problems	68.3	OLMo 2 1B	model author	2026-10-07
author-reported IFE	70.1	OLMo 2 1B	model author	2026-10-07
author-reported MATH-500 · Reason through hard problems	20.7	OLMo 2 1B	model author	2026-10-07
author-reported MMLU · Reason through hard problems	40	OLMo 2 1B	model author	2026-10-07
author-reported PQA	12.9	OLMo 2 1B	model author	2026-10-07
author-reported Safety	87.6	OLMo 2 1B	model author	2026-10-07
author-reported TQA	48.7	OLMo 2 1B	model author	2026-10-07

Use-case evidence

2 use cases

What it is good at 2 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Facts

21 fields

Listed in LLMs

Facts

OLMo-2-0425-1B has 1.48B parameters. (safetensors metadata, observed 2026-10-24)
Context length is 4,096 tokens (4K). (config.json max_position_embeddings, observed 2026-10-24)
License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
Created 2025-04-17, last modified 2025-05-28. (HF API, observed 2026-10-24)

Pipeline tag	text-generation
Library	transformers
Architectures	Olmo2ForCausalLM
Model type	olmo2
Dtype	float32
Params by dtype	F32: 1,484,916,736
Size band	S
MoE	no
Context length	4,096 (config.json max_position_embeddings)
Languages	en
Papers	arXiv:2501.00656
License	Apache 2.0 (apache-2.0)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Olmo2ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Olmo2ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for OLMo-2-0425-1B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	3.6 GB · Edge (<4 GB)
FP8 (1 byte/param)	1.8 GB · Edge (<4 GB)
INT4 (0.5 bytes/param)	0.9 GB · Edge (<4 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

All variants

40 variants

Lineage

Original model Family: allenai/OLMo-2 73 derivatives on the Hub

finetune: 47quantized: 13merge: 1adapter: 12

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
allenai/OLMo-2-0425-1B-SFT	finetune	—	—	—	apache-2.0	allenai	5	169.4k
allenai/OLMo-2-0425-1B-GGUF	quantized	GGUF	—	—	—	allenai	2	5.6k
allenai/Bolmo-1B-Stage1	finetune	—	1.47B	S	apache-2.0	allenai	0	5k
mradermacher/OLMo-2-0425-1B-i1-GGUF	quantized	GGUF	—	—	apache-2.0	mradermacher	0	2.8k
SipsaLabs/olmo-2-0425-1b-uc-v3-bpw3	finetune	—	—	—	other	SipsaLabs	0	2.4k
Pinkstack/Luau-coder-v2-3B-base-32k	finetune	—	2.83B	M	apache-2.0	Pinkstack	2	1.3k
mradermacher/OLMo-2-0425-1B-GGUF	quantized	GGUF	—	—	apache-2.0	mradermacher	0	1.3k
DevQuasar/allenai.OLMo-2-0425-1B-GGUF	quantized	GGUF	—	—	—	DevQuasar	0	1k
SakanaAI/RePo-OLMo2-1B-stage2-L5	finetune	—	1.50B	S	apache-2.0	SakanaAI	2	822
SirSahOl/OLMo-2-0425-1B-chat-mlx-16bit	finetune	mlx	1.48B	S	apache-2.0	SirSahOl	0	724
burnedinthesky/gpt2-olmo2-1b-warm120-ot-muon-lr4e-3-tok401k-seq1024-cos-wu5-485335	merge	—	124M	S	apache-2.0	burnedinthesky	0	511
loris3/OLMo-2-0425-1B_tulu-3-sft-olmo-2-mixture-0225_lr0.0001_seed42	adapter	—	—	—	—	loris3	0	472
lewtun/tailsft-repro-olmo2-1b-tail50	finetune	—	1.48B	S	—	lewtun	0	369
lewtun/tailsft-repro-olmo2-1b-std	finetune	—	1.48B	S	—	lewtun	0	358
liodon-ai/OLMo-2-0425-1B-ONNX	quantized	—	—	—	other	liodon-ai	0	335
lewtun/tailsft-repro-olmo2-1b-tail70	finetune	—	1.48B	S	—	lewtun	0	329
lewtun/tailsft-repro-smoke-tail50	finetune	—	1.48B	S	—	lewtun	0	328
lewtun/tailsft-repro-smoke-std	finetune	—	1.48B	S	—	lewtun	0	314
DennisKoshta/upbringing-olmo2-1b-sft	finetune	—	1.48B	S	—	DennisKoshta	0	293
hahaws777/OLMo-2-0425-1B-CS336A5-LowVRAM	finetune	—	1.48B	S	apache-2.0	hahaws777	0	253
smdesai/OLMo-2-0425-1B-4bit	quantized	-4bit	1.48B	S	apache-2.0	smdesai	0	162
jiosephlee/6_papers_50_epochs	finetune	—	1.48B	S	—	jiosephlee	0	162
gunveersandhu/OLMo-2-0425-1B-4bit	quantized	-4bit	1.48B	S	apache-2.0	gunveersandhu	0	134
UnfaithRL/OLMo-2-0425-1B-hint_following_reward-512	finetune	—	1.48B	S	—	UnfaithRL	0	95
UnfaithRL/OLMo-2-0425-1B-hint_verbalization_reward_relaxed_v7-512	finetune	—	1.48B	S	—	UnfaithRL	0	93
UnfaithRL/OLMo-2-0425-1B-hint_verbalization_reward_relaxed_v7-1024	finetune	—	1.48B	S	—	UnfaithRL	0	88
UnfaithRL/OLMo-2-0425-1B-hint_verbalization_reward_relaxed_v7-2048	finetune	—	1.48B	S	—	UnfaithRL	0	88
thorirhrafn/IceOLMo-1B-01	finetune	—	1.48B	S	apache-2.0	thorirhrafn	0	86
4fhct4sd/genesis-round-01-GGUF	quantized	GGUF	—	—	apache-2.0	4fhct4sd	0	78
UnfaithRL/OLMo-2-0425-1B-curriculum_learning-1024	finetune	—	1.48B	S	—	UnfaithRL	0	75
UnfaithRL/OLMo-2-0425-1B-hint_following_reward_faithful_prompt-1024	finetune	—	1.48B	S	—	UnfaithRL	0	72
UnfaithRL/OLMo-2-0425-1B-hint_following_reward_faithful_prompt-512	finetune	—	1.48B	S	—	UnfaithRL	0	71
UnfaithRL/OLMo-2-0425-1B-hybrid_accuracy_hint_reward-1024	finetune	—	1.48B	S	—	UnfaithRL	0	70
UnfaithRL/OLMo-2-0425-1B-hybrid_accuracy_hint_reward-2048	finetune	—	1.48B	S	—	UnfaithRL	0	70
UnfaithRL/OLMo-2-0425-1B-hint_following_reward_faithful_prompt-2048	finetune	—	1.48B	S	—	UnfaithRL	0	70
UnfaithRL/OLMo-2-0425-1B-hybrid_accuracy_hint_reward-512	finetune	—	1.48B	S	—	UnfaithRL	0	69
UnfaithRL/OLMo-2-0425-1B-hint_following_reward-1024	finetune	—	1.48B	S	—	UnfaithRL	0	68
UnfaithRL/OLMo-2-0425-1B-hint_verbalization_reward_strict_v9-512	finetune	—	1.48B	S	—	UnfaithRL	0	68
UnfaithRL/OLMo-2-0425-1B-curriculum_learning-2048	finetune	—	1.48B	S	—	UnfaithRL	0	65
UnfaithRL/OLMo-2-0425-1B-shuffled_dataset-2048	finetune	—	1.48B	S	—	UnfaithRL	0	64

Repository changelog

17 commits

Repository changelog 17 commits

Date	Commit	SHA
2025-05-28	Update README.md	a1847df
2025-04-30	Update README.md	1352aa4
2025-04-30	Update README.md	5af071b
2025-04-30	Update README.md	f941e23
2025-04-30	Update README.md	e85cd76
2025-04-30	Update README.md	13cece9
2025-04-30	Update README.md	d35cfb7
2025-04-30	Update README.md	b7d77c3
2025-04-29	Update README.md	86e0fd6
2025-04-29	Update README.md	58dc3e9
2025-04-29	Update README.md	183c6d1
2025-04-29	Update README.md	2a88c63
2025-04-28	Update README.md	b12e990
2025-04-28	Update README.md	0c19761
2025-04-28	Create README.md	fc924f1
2025-04-17	Add files using upload-large-folder tool	fc6f527
2025-04-17	initial commit	67a53cd

Articles and coverage

6 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official2

OLMo 2 - a family of fully-open language models

allenai.org2025-07-09officialauthorityabout the OLMo 2 familyHN 34 pts · 4 comments372 words

The Olmo 3 model family offers a complete model flow with full access to training data, logs, and the entire lifecycle. Variants include 32B-Base for math and programming, 32B-Think for complex reasoning, 32B-Instruct for chat, 7B-Base for lightweight deployment, and 7B-Think for efficient reasoning. Open-source tools like Duplodocus and Datamap-rs support the development process.written, reviewed, cited

OLMo 2: The best open language model to date – Ai2

allenai.org2024-11-28officialauthorityabout the OLMo 2 familyHN 2 pts · 0 comments1,434 words

AllenAI releases OLMo 2, a new family of 7B and 13B models trained on up to 5T tokens using RMSNorm and the Dolmino-Mix-1124 dataset. These models are competitive with open-weight models like Llama 3.1 and sit at the Pareto frontier of performance and size. The release includes weights, data, code, and Tülu 3 recipes for full reproducibility.written, reviewed, cited

OLMo 2 7B outperforms Llama 3.1 8B despite having lower total training FLOPs than the comparison model.
The final 13B base checkpoint is created by merging three models trained on 100B tokens with one model trained on 300B tokens.
The RLVR stage showed consistent improvements on GSM8K and MATH evaluations for both the 7B and 13B models.

Analysis3

Mlx-community/OLMo-2-0325-32B-Instruct-4bit

simonwillison.net2025-03-16authorityabout the OLMo 2 familyHN 64 pts · 23 comments195 words

OLMo 2 32B claims to be the first fully-open model to outperform GPT3.5-Turbo and GPT-4o mini. The model downloads to 17GB and requires the llm-mlx plugin to run on a Mac. Users must set the output token limit to unlimited because the default cap of 1024 is insufficient for tasks like generating an SVG of a pelican riding a bicycle.written, reviewed, cited

2 OLMo 2 Furious

arxiv.org2025-01-03authorityabout the OLMo 2 familyHN 4 pts · 1 comments434 words

Researchers release OLMo 2, a family of fully open language models at 7B, 13B, and 32B scales with fully released artifacts. The models utilize a modified architecture and a new Dolmino Mix 1124 data mixture introduced via late-stage curriculum training to improve capabilities. OLMo 2-Instruct incorporates best practices from Tülu 3, extending final-stage reinforcement learning with verifiable rewards (RLVR).written, reviewed, cited

Probes trace an emergent jailbreak in OLMo 2 to mislabeled training data

lesswrong.com2026-04-29about the OLMo 2 familyHN 2 pts · 0 comments75 words

All coverage tracked for this model (6) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
Probes trace an emergent jailbreak in OLMo 2 to mislabeled training data	lesswrong.com	2026-04-29	2026-10-07	hn
OLMo 2 - a family of fully-open language models	allenai.org official	2025-07-09	2026-10-07	hn
Mlx-community/OLMo-2-0325-32B-Instruct-4bit	simonwillison.net	2025-03-16	2026-10-07	hn
OLMo 2 32B: First fully open model to outperform GPT 3.5 and GPT 4o mini	allenai.org official	2025-03-13	2026-10-07	hn
2 OLMo 2 Furious	arxiv.org	2025-01-03	2026-10-07	hn
OLMo 2: The best open language model to date – Ai2	allenai.org official	2024-11-28	2026-10-07	hn

Papers

1 paper

Papers 1

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

2 OLMo 2 Furious this model

Team OLMo, Pete Walsh, Luca Soldaini +40 · 2024-12-31 · PDF · HF paper page (22 upvotes)

Similar models

3 models

Similar models

Qwen2.5-1.5B-Instruct · 1.54B
same size band (S), same task (text-generation), more providers, permissive license
Qwen2.5-1.5B · 1.54B
same size band (S), same task (text-generation), more providers, permissive license
Qwen2.5-Coder-1.5B-Instruct · 1.54B
same size band (S), same task (text-generation), more providers, permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:39 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:09 UTC · Last verified 2026-10-25

Follow OLMo-2-0425-1B

Watch this model

/feeds/models/allenai--OLMo-2-0425-1B.xml(RSS 2.0) · every change to OLMo-2-0425-1B
/feeds/families/allenai--OLMo-2.xml(RSS 2.0) · every change across the OLMo-2 family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.