2 OLMo 2 Furious
Summary
OLMo 2 is the next generation of the authors' fully open language models, a family of dense autoregressive models at 7B, 13B, and 32B scales, with released weights, full training data, training code and recipes, logs, and thousands of intermediate checkpoints. The updated pretraining mixture introduces Dolmino Mix 1124. They incorporate practices from Tulu 3 to develop OLMo 2-Instruct, including reinforcement learning with verifiable rewards (RLVR). [S1]
Claims
- id: C1 claim: 'OLMo 2 includes dense models at 7B, 13B, and 32B with fully released training artifacts.' source: S1 evidence: "dense autoregressive language models at 7B, 13B and 32B scales with fully released artifacts"
- id: C2 claim: 'The updated data mixture introduces Dolmino Mix 1124.' source: S1 evidence: "a new, specialized data mix called Dolmino Mix 1124"
- id: C3 claim: 'OLMo 2-Instruct uses reinforcement learning with verifiable rewards (RLVR).' source: S1 evidence: "final-stage reinforcement learning with verifiable rewards (RLVR)"
Verification
- Code available: Comments point to a model demo at https://playground.allenai.org, which returned HTTP 200 on 2026-11-12. The HTML also points to an initial-release post at https://allenai.org/blog/olmo2, which returned HTTP 200. Not run.
- Reproduced: not reproduced (2026-11-12). Code not run.
- Related work: This batch also has the original OLMo note. Live note 2302-13971-llama.
- Limits: This note follows the v3 text last revised 8 Oct 2025, updated to include 32B results. Pareto-frontier comparisons are author-reported.
- License: CC BY 4.0 on the abs page (2026-11-12; icon by-4.0.png; http://creativecommons.org/licenses/by/4.0/). Submitted 31 Dec 2024 (v1); last revised 8 Oct 2025 (v3). Sources: [S1] https://arxiv.org/abs/2501.00656, v3 (8 Oct 2025). checked_on: 2026-11-12.
When to apply
Use when asking how OLMo 2 extended a fully open training recipe, including Dolmino Mix 1124 and RLVR.
Facts
- id: F1 claim: 'OLMo 2 is a fully open dense language-model family at 7B, 13B, and 32B scales.' confidence: medium date: 2026-11-12 sources: [S1] license: CC BY 4.0
- id: F2 claim: 'The authors introduce Dolmino Mix 1124 in late-stage curriculum pretraining.' confidence: medium date: 2026-11-12 sources: [S1] license: CC BY 4.0
- id: F3 claim: 'OLMo 2-Instruct incorporates Tulu 3 practices, including reinforcement learning with verifiable rewards.' confidence: medium date: 2026-11-12 sources: [S1] license: CC BY 4.0
Sources
- [S1] arXiv 2501.00656, https://arxiv.org/abs/2501.00656. License: CC BY 4.0. v3 (8 Oct 2025). Accessed 2026-11-12.
License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted). Paper is CC BY 4.0.
License: CC-BY-4.0 · Summary by iinventedlight (AI-assisted). Source: Team OLMo and 42 coauthors (listed on the arXiv abs page), 2 OLMo 2 Furious (CC BY 4.0).
Cite this
'iinventedlight (AI-assisted)'. 2026. "2 OLMo 2 Furious." iinventedlight Library, iil:note:6bs41svs, version of 2026-11-12. https://iinventedlight.com/library/notes/2501-00656-olmo-2/. License: CC-BY-4.0.
@misc{iil_note_6bs41svs,
author = {{'iinventedlight (AI-assisted)'}},
title = {{2 OLMo 2 Furious}},
howpublished = {iinventedlight Library},
year = {2026},
date = {2026-11-12},
url = {https://iinventedlight.com/library/notes/2501-00656-olmo-2/},
note = {Permanent ID iil:note:6bs41svs. License: CC-BY-4.0.},
}
Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.
For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2501-00656-olmo-2.md.