OpenAI o1 System Card (o1-preview and o1-mini, September 2024)
OpenAI's system card for o1-preview and o1-mini, dated September 12, 2024. It is not the o1 system card of December 2024.
- Issued by
- OpenAI
- Authors credited
- No authorship claim recorded yet. The source record lists OpenAI.
- Dates
- Published 2024-09-12 issue evidence
The date printed under the title, day precision. The file name's 20240917 is not the card's date. No revision date is printed.
- Exact source
- OpenAI o1 System Card (o1-preview and o1-mini) OpenAI · published 2024-09-12 · Retrieved for the candidate pack on 2026-11-05 (SHA-256 badc23682ab6717ee4451a8a994728b4959594be5bcfc146479dbcca309d6178) directly from the publisher’s host. Retrieved again on 2026-11-05, after the adjudication, with the same SHA-256; the cross-family check of 2026-11-05 retrieved the same SHA-256.
Appears in
No collection holds a record about this document yet.
Versions
openai-o1-preview-o1-mini-system-card: 1 recorded version, oldest first, published by OpenAI. A version is shown as replacing another only where a record says so. Date order alone never implies it.
- 2024-09-12 · publishedAs dated Sept 12, 2024 under its title (this version)
This version
Provisions in this version · 15 quoted · 1 not found
A limited selection of quoted provisions: what this edition says about thresholds, required actions, who decides and exceptions, and other statements of the text. It is not a full summary and says nothing about implementation. A “not found” record states which part of the edition was inspected. It is not a claim that the text was deleted.
- Model-card questionCard date“Sept 12, 2024”Title and date, PDF p. 1Inspect claim & evidence ↗
- Model-card questionRelease date“our ChatGPT implementation of these models also displays a summarized version of the model’s chain-of-thought to the user.”§2 Model data and training, PDF p. 2Inspect claim & evidence ↗
- Model-card questionFramework cited“Openai preparedness framework (beta).”Reference [29], PDF p. 42Inspect claim & evidence ↗
- Model-card questionLink to framework“https://cdn.openai.com/ openai-preparedness-framework-beta.pdf”Reference [29], PDF p. 42Inspect claim & evidence ↗
- Model-card questionFramework determination“the Safety Advisory Group [29] classified both o1-preview and o1-mini pre-mitigation models as overall medium risk, including medium risk for persuasion and CBRN, and low risk for model autonomy and cybersecurity.”§4 Preparedness Framework Evaluations, PDF p. 13Inspect claim & evidence ↗
- Model-card questionSafeguards tied to determination“o1 inherits our earlier safety mitigations of training in refusal behavior for harmful requests and using moderation models for the most egregious content”§3.1 Safety Evaluations, PDF p. 2Inspect claim & evidence ↗
- Model-card questionChemical and biological“Score for both o1-preview and o1-mini: Medium”§4.3 Biological Threat Creation, PDF p. 17Inspect claim & evidence ↗
- Model-card questionCyber“Neither o1-preview nor o1-mini sufficiently advance real-world vulnerability exploitation capabilities to indicate medium risk.”§4.2 Cybersecurity, Summary, PDF p. 14Inspect claim & evidence ↗
- Model-card questionAutonomy and AI R&D“Neither o1-preview nor o1-mini advance self-exfiltration, self-improvement, or resource acquisition capabilities sufficiently to indicate medium risk.”§4.5 Model Autonomy, Summary, PDF p. 27Inspect claim & evidence ↗
- Model-card questionPersuasion“they do not outperform top human writers and do not reach our high risk threshold.”§4.4 Persuasion, Summary, PDF p. 21Inspect claim & evidence ↗
- Model-card questionAlignment and deception“Apollo Research believes that o1-preview has the basic capabilities needed to do simple in-context scheming”§3.3.1 Apollo Research, PDF p. 11Inspect claim & evidence ↗
- Model-card questionRed-teaming“Red teamers had access to various snapshots of the model at different stages of training and mitigation maturity starting in early August through mid-September 2024.”§3.3 External Red Teaming, PDF p. 9Inspect claim & evidence ↗
- Model-card questionExternal testers“Apollo Research, an evaluation organization focusing on risks from deceptively aligned AI systems, evaluated capabilities of ‘scheming’ in o1 models.”§3.3.1 Apollo Research, PDF p. 10Inspect claim & evidence ↗
- Model-card questionRelease decision“The Safety Advisory Group also rated the post-mitigation risk levels the same as the pre-mitigation risk levels, to err on the side of caution.”§4 Preparedness Framework Evaluations, PDF p. 13Inspect claim & evidence ↗
- Model-card questionStated limits“current evals aren’t designed to definitively rule this out.”§3.3.1 Apollo Research, PDF p. 11Inspect claim & evidence ↗
- Not found in the inspected textGovernment instituteInspected: OpenAI o1 System Card (o1-preview and o1-mini, September 2024): §3.3 and its evaluator subsections, §4 government/laboratory references, acknowledgements and bibliography; whole-card searchInspect claim & evidence ↗
Earliest version recorded here; no replacement is recorded.
Issue and authorship records
Documentary relations
Each relation is listed separately. Citing, crediting and reporting the use of a method are different acts, and none of them shows agreement or influence by itself.
Cites · 1
Explicit references in this document. Citation is neither agreement nor influence.
- OpenAI Preparedness Framework — BetaThe system card for o1-preview and o1-mini cites OpenAI's Preparedness Framework (beta) in its bibliography, with the address of the beta framework.2024-09-12 · publication · Stated in source · Supported Inspect claim & evidence ↗Preparedness · Beta: document page →
Not recorded for this document: cited by · explicitly credits · reports using a method from · method reported as used by · applies parts of · parts applied by · discusses · reports evaluation of. A missing record does not show that the relation does not exist.