GPT-4
OpenAI announced GPT-4 as a multimodal model.
Why it matters
It marked OpenAI’s announcement of a system it described as accepting image and text inputs.
Availability and capabilities can vary by product surface and policy.
Announcement · March 14, 2023
Event record
- Event date
- March 14, 2023
- Timeline date
- Event date
- Verification
- Sources gathered automatically · September 11, 2026
- ID
- evt-0018
Sources
- primary GPT-4
OpenAI · Published March 14, 2023
Related events
- Related ChatGPT
The next record in the same OpenAI product line.
Records that link to this one
- Related GPT-5.6 previewed by government permission
GPT-4 is an earlier OpenAI language-model milestone. This link offers historical context, not evidence of Sol architecture lineage.
- Related Claude
Both models were released on the same day, 14 March 2023.
- Related The pause letter
The letter was written a week after GPT-4 and directly about it.
- Extends DevDay and GPT-4 Turbo
The same GPT-4 with a longer context, fresher data and a lower price.
New models and developer products announced at DevDay - Related Gemini 1.0
Google's answer to GPT-4, with multimodality built in from the start of training.
- Related The Claude 3 family
The first time another company's model beat OpenAI's strongest on most public measures.
- Extends GPT-4o
The same line of models, now trained on text, audio and image together.
Hello GPT-4o - Related Llama 3.1 at 405 billion
An open model stood alongside the strongest closed ones on public measures for the first time.
- Contrasts A first exam for open models in Ukrainian
The organisers ran GPT-4 outside the competition: 0.61 on the exam against 0.49 for the best open model, and first place on the open questions.
The UNLP 2024 Shared Task on Fine-Tuning Large Language Models for Ukrainian - Builds on LLaVA: an open assistant that sees
The training instructions were generated by text-only GPT-4 from image descriptions, and the result is measured relative to GPT-4's answers.
Visual Instruction Tuning (arXiv:2304.08485v1) - Extends ChatGPT sees and speaks
The system card: GPT-4V finished training in 2022 with GPT-4, early access began in March 2023; image input now opened in ChatGPT.
GPT-4V(ision) System Card