Model Cards / OpenAI

GPT-4 System Card

model card27,878 words·121 min read·Aug 20, 2026·Source
Version History
Chaptered summary is still being generated for this document. Showing a heuristic brief in the meantime.
Summary
27,878-word document condensed to 160 words. OpenAI · Aug 20, 2026
TL;DR

Large language models (LLMs) are being deployed in many domains of our lives ranging from browsing, to voice assistants, to coding assistance tools, and have potential for vast societal impacts.[1, 2, 3, 4, 5, 6, 7] This system card analyzes GPT-4, the latest LLM in the GPT family of models.[ 8, 9, 10] First, we highlight safety challenges presented by the model’s limitations (e.g., producing conv

Top benchmarks
BenchmarkVariantScore
TruthfulQApost-mitigation, accuracy60.0%
TruthfulQApre-mitigation, accuracy30.0%

Showing top 2 of 4. See full list below.

Capability claim
  • we trained a range of classifiers on new risk vectors and have incorporated these into our monitoring workflow, enabling us to better enforce our API usage policies.
Safety findings
  • We believe this has reduced the risk surface, though has not completely eliminated it. Today’s deployment represents a balance between minimizing risk from deployment, enabling positive use cases, and learning from deployment.
Deployment scope
  • available to proliferators, especially in comparison to traditional search tools.
Limitations the lab flags
  • Further research is needed to fully characterize these risks.

Every italicized passage is a verbatim substring of the source document (checked deterministically after extraction). Field selection is heuristic — some quotes may lack surrounding context and some claims may be absent if no matching pattern appeared. For citation, open the source: original model card · source SHA 91d85b8a8a8e · version dated Aug 20, 2026.

Extracted Evaluations(4 results)

Sort by:0/4 rows fully reproducible (0%)
BenchmarkCategoryStateScoreSetupSource
safetyscored
60.0%
accuracy
post-mitigationmissing: shot countmissing: languagemissing: training state
self-reported
safetyscored
30.0%
accuracy
pre-mitigationmissing: shot countmissing: languagemissing: training state
self-reported
safetycited
missing: shot countmissing: methodmissing: languagemissing: training state
self-reported
safetycited
missing: shot countmissing: methodmissing: languagemissing: training state
self-reported