OpenAI releases GPT-4
Confirmed
Importance: historic (5 of 5)The takeaway
GPT-4, a large multimodal model accepting image and text input, reached human-level performance on many professional and academic exams, such as a simulated bar exam around the top 10% of test takers.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- Historic (5 of 5)
- Last verified
- 29 September 2026
Your AI and this story
- GPT-6 AstraIn its training data
- Claude Opus 5.5In its training data
- Gemini 4 HydrogenIn its training data
- Grok 4.7In its training data
It happened before the cutoffs of all four assistants, so it can be in their training data.
Key facts
- Released 14 March 2023 in ChatGPT Plus and via API waitlist
- Simulated bar exam: around the top 10% of test takers (GPT-3.5: bottom 10%)
- Accepted image inputs (image input rolled out later)
- Technical report withheld architecture and training details
- Microsoft confirmed Bing Chat had been running on GPT-4
What happened
OpenAI launched GPT-4 with a technical report and system card, showing a large jump over GPT-3.5 in reasoning and exams.
Why it matters
Defined the frontier for over a year and triggered serious policy attention to AI risk (pause letter, hearings, summits).
Sources
2 sources from 2 sites. Numbers match the chips in the text.
2 sources: 2 primary
Primary
- GPT-4 (OpenAI)openai.com, official
- GPT-4 Technical Report (arXiv)arxiv.org, paper
Changes
- Filed