Model components / Base / pretrained models

GPT-4o

OpenAI's legacy multimodal foundation model (May 2024) that natively handled text, image, and audio. Now two numerical generations behind: superseded by the GPT-5 line (GPT-5 in August 2025, then 5.2, 5.3, 5.4, and GPT-5.5 on April 23, 2026, the first fully retrained base since GPT-4.5). All access remains API-only via post-trained variants; GPT-4o is being sunset as the ChatGPT default in favor of GPT-5.5 Instant (May 5, 2026). Listed as the historical closed-source baseline that open-weights models were measured against through 2024-2025.

OpenAI GPT-4o, multimodal (text+image in, text out), 128K context; first snapshot gpt-4o-2024-05-13, knowledge cutoff Oct 2023. API-only, no released weights. Verified live in OpenAI's model docs June 2026 (developers.openai.com). A 2024-era flagship, since surpassed by the GPT-5 line but still served via API/ChatGPT.

Openness

1 high confidence

1.0

weights: closed
data: closed
code: closed
license: Proprietary(API-only)

No downloadable weights; API/ChatGPT-only, like the GPT-5 anchor (O1 closed). Confirmed API-only with 128K context and Oct 2023 cutoff on OpenAI docs 2026-06-04.

https://developers.openai.com/api/docs/models/gpt-4o recorded 2026-06-04
GPT-4o API-only model, 128K context, text+image input, no open weights, Oct 2023 cutoff

Adoption

5 medium confidence

5.0

GPT-4o was the default ChatGPT model for much of 2024-2025 before GPT-5; ChatGPT crossed ~900M WAU (Feb 2026) and ~1B (May 2026). Attributed honestly to the ChatGPT surface GPT-4o powered, not a standalone GPT-4o count; well above the >10M threshold. WAU figure corroborated via aggregator (demandsage), not a primary OpenAI disclosure, but the threshold is not close so no downgrade.

https://developers.openai.com/api/docs/models/gpt-4o recorded 2026-06-04
GPT-4o served as a primary OpenAI API/ChatGPT model
https://www.demandsage.com/chatgpt-statistics/ recorded 2026-06-04
ChatGPT ~900M+ weekly active users (aggregator, corroborating surface scale)

Capability

3 medium confidence

3.0

Frontier in 2024 but surpassed by 2026 reasoning frontier (GPT-5 C5, Gemini 3.5 C5); per the recipe, a mid-2025-and-earlier model that has been surpassed scores 3, not 5. Caveat: the cited OpenAI docs page (verified 2026-06-04) shows NO benchmark figures; it only calls GPT-4o 'our most capable model outside our o-series'; the MMLU/MMMU values are launch-era external numbers and need a primary benchmark source to back the headline.

https://developers.openai.com/api/docs/models/gpt-4o recorded 2026-06-04
GPT-4o positioned as 'most capable model outside o-series', 128K context, text+image; no benchmark figures present

Unchanged since 2026-07-30 (last edited, not re-checked)