GPT-4o

weights

closed

data

closed

code

closed

license

Proprietary(API-only)

GPT-4o is available only through the OpenAI API, with no downloadable weights. OpenAI publishes neither its training data nor its training code.

• https://artificialanalysis.ai/models/gpt-4o recorded 2026-09-29

  Artificial Analysis states "GPT-4o (Nov '24) is a proprietary model and OpenAI has not disclosed the model size or parameter count" and measures it only on the vendor endpoint - "generates output at 167.9 tokens per second (based on OpenAI"; no open-weights provider is listed.

• https://developers.openai.com/api/docs/models/gpt-4o recorded 2026-09-29

  OpenAI API model page describing GPT-4o as "is our versatile, high-intelligence flagship model" that "accepts both text and image inputs, and produces text outputs", sold per token ("Pricing is based on the number of tokens used") and pinned through API snapshots ("Snapshots let you lock in a specific version of the model"). Searched the body for weights, download, open-weight and Hugging Face: none, and no training corpus or training code is published.

GPT-4o was the default ChatGPT model through most of 2024 and 2025 before GPT-5 replaced it, so adoption here is read from ChatGPT's own weekly active users rather than from a standalone GPT-4o count that OpenAI has never published. Those weekly-actives figures come from OpenAI's own announcements rather than from an aggregator's estimate.

• https://almcorp.com/blog/chatgpt-900-million-weekly-active-users/ recorded 2026-08-14

  "On February 27, 2026, OpenAI announced that ChatGPT had surpassed 900 million weekly active users" - the surface figure this reading is based on

• https://developers.openai.com/api/docs/models/gpt-4o recorded 2026-08-13

  GPT-4o served as a primary OpenAI API/ChatGPT model; no usage figure published on the page

• https://www.demandsage.com/chatgpt-statistics/ recorded 2026-06-04

  ChatGPT ~900M+ weekly active users (aggregator, corroborating surface scale). Not re-fetchable - the host answers HTTP 403 to consecutive requests - so it keeps its earlier read.

GPT-4o was OpenAI's multimodal flagship in 2024 and now sits well below a frontier that includes OpenAI's own GPT-5 and GPT-6 lines. On Artificial Analysis's Intelligence Index it scores below the median for non-reasoning models in its price tier, and OpenAI's model page publishes no benchmark figures.

• https://artificialanalysis.ai/models/gpt-4o recorded 2026-09-29

  Artificial Analysis model page for the November 2024 snapshot - "GPT-4o (Nov '24) scores 8 (estimated) on the Artificial Analysis Intelligence Index, placing it below average among other non-reasoning models in a similar price tier (median: 15)", on "Artificial Analysis Intelligence Index v4.3.2".

• https://developers.openai.com/api/docs/models/gpt-4o recorded 2026-09-29

  OpenAI API model page for GPT-4o with "128,000 context window" and "Oct 01, 2023 knowledge cutoff", described as "our most capable model outside of our o-series models"; the page carries no benchmark figures.

• https://en.wikipedia.org/wiki/GPT-4o recorded 2026-09-29

  Encyclopedia entry recording the launch figure "GPT-4o scored 88.7 on the Massive Multitask Language Understanding" benchmark, against 86.5 for GPT-4.