llm-architectures-explained

/models

GPT-4

OpenAI · GPT-4 · closed weights

Facts and where they come from

Released2023-03labofficial docssnapshot gpt-4-0314
Licenceproprietarylabofficial docsweights not published
Total parametersnot disclosednot disclosed
Active parametersnot disclosednot disclosed
Context length8K tokenslabofficial docsmodel page: 8,192 context window

Reported estimates

OpenAI has not disclosed these. They are third-party estimates, shown only with their label, source, date and confidence, and never used by the calculator unless you switch estimates on.

  • Total parameters:reported estimate ≈1.76T (“GPT-4 (≈1.76T)”)Source: Ben Abacha, Yim, Fu, Sun, Yetisgen, Xia and Lin, MEDEC (arXiv 2412.19260v1), 2024-12-26. §4 (p. 6): model sizes listed in the experiments sectionConfidence: Low. The authors write that the sizes are estimates mined from public articles, that they cannot vouch for their accuracy, and that they are given only to help place the results.

Architecture

not disclosed No diagram or modelled costs: the GPT-4 Technical Report states that it gives no details of the architecture or model size.

Sources