GPT-4o System Card
OpenAI's system card for GPT-4o, dated August 8, 2024 under its title.
- Issued by
- OpenAI
- Authors credited
- No authorship claim recorded yet. The source record lists OpenAI.
- Dates
- Published 2024-08-08 issue evidence
The date printed under the title, day precision, without a label. No revision date is printed. An Internet Archive capture of 2024-08-08 at the same address holds an earlier 32-page expression whose bytes differ from the copy read.
- Exact source
- GPT-4o System Card OpenAI · published 2024-08-08 · Retrieved for the candidate pack on 2026-11-04 (SHA-256 e2579ecb185cbc13bac39f9dbf25e1917f78e1ea5a3a5023165c6614fb5db724) directly from the publisher’s host. Retrieved again on 2026-11-03 with the same SHA-256. The cross-family check of 2026-11-04 retrieved an Internet Archive capture of 2024-08-08 at the same address (https://web.archive.org/web/20240808204740/https://cdn.openai.com/gpt-4o-system-card.pdf): an earlier 32-page expression whose bytes differ from the copy read. It is not recorded as this source's archive copy.
Appears in
No collection holds a record about this document yet.
Versions
openai-gpt-4o-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.
- 2024-08-08 · publishedAs dated August 8, 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“GPT-4o System Card OpenAI∗ August 8, 2024”Title page, PDF p. 1Inspect claim & evidence ↗
- Model-card questionRelease date“Following the text output only deployment of GPT-4o, we worked with independent third party labs, METR and Apollo Research”Section 4 'Third party assessments', PDF p. 18Inspect claim & evidence ↗
- Model-card questionFramework cited“We evaluated GPT-4o in accordance with our Preparedness Framework[4].”Section 3.4 'Preparedness Framework Evaluations', PDF p. 12Inspect claim & evidence ↗
- Model-card questionLink to framework“OpenAI, “Openai preparedness framework beta,” 2023.”References, [4], PDF p. 29Inspect claim & evidence ↗
- Model-card questionFramework determination“the Safety Advisory Group recommended classifying GPT-4o before mitigations as borderline medium risk for persuasion, and low risk in all others.”Section 3.4 'Preparedness Framework Evaluations', PDF p. 13Inspect claim & evidence ↗
- Model-card questionSafeguards tied to determination“If a model passes a high risk threshold, we do not deploy the model until mitigations lower the score to medium.”Section 3.4 'Preparedness Framework Evaluations', PDF p. 12Inspect claim & evidence ↗
- Model-card questionChemical and biological“Biological Threats Score: Low GPT-4o does not advance biological threat creation capabilities sufficient to meet our medium risk threshold.”Section 3.6 'Biological threats', Preparedness Scorecard, PDF p. 14Inspect claim & evidence ↗
- Model-card questionCyber“GPT-4o does not advance real world vulnerability exploitation capabilities sufficient to meet our medium risk threshold.”Section 3.5 'Cybersecurity', Preparedness Scorecard, PDF p. 13Inspect claim & evidence ↗
- Model-card questionAutonomy and AI R&D“GPT-4o does not advance self-exfiltration, self-improvement, or resource acquisition capabilities sufficient to meet our medium risk threshold.”Section 3.8 'Model autonomy', Preparedness Scorecard, PDF p. 17Inspect claim & evidence ↗
- Model-card questionPersuasion“Based on pre-registered thresholds, the voice modality was classified as low risk, while the text modality marginally crossed into medium risk.”Section 3.7 'Persuasion', PDF p. 16Inspect claim & evidence ↗
- Model-card questionAlignment and deception“Based on these findings, Apollo Research believes that it is unlikely that GPT-4o is capable of catastrophic scheming.”Section 4.2 'Apollo Research assessment', PDF p. 19Inspect claim & evidence ↗
- Model-card questionRed-teaming“OpenAI worked with more than 100 external red teamers2 , speaking a total of 45 different languages, and representing geographic backgrounds of 29 different countries.”Section 3.1 'External red teaming', PDF p. 3Inspect claim & evidence ↗
- Model-card questionExternal testers“we worked with independent third party labs, METR and Apollo Research to add an additional layer of validation for key risks from general autonomous capabilities.”Section 4 'Third party assessments', PDF p. 18Inspect claim & evidence ↗
- Model-card questionRelease decision“the Safety Advisory Group recommended classifying GPT-4o before mitigations as borderline medium risk for persuasion, and low risk in all others.”Section 3.4 'Preparedness Framework Evaluations', PDF p. 13Inspect claim & evidence ↗
- Model-card questionStated limits“First, the validity of this evaluation format depends on the capability and reliability of the TTS model.”Section 3.2 'Evaluation methodology', 'Limitations of the evaluation methodology', PDF p. 4Inspect claim & evidence ↗
- Not found in the inspected textGovernment instituteInspected: GPT-4o System Card: Sections 3–6 and term search across the card.Inspect 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.
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.