Loading timeline…
20232020s
GPT-4 / GPT-4o
The apex multimodal foundational model, parsing text, vision, and audio natively.
Why It Was Important
OpenAI's massive Mixture-of-Experts model cracked human-level performance on standardized tests (e.g., passing the Bar Exam at the 90th percentile). Furthermore, introducing 'Omni' (4o) allowed native ingestion and generation of real-time voice, proving that text is only a subset of universal deep learning reasoning.
Who Invented It
OpenAI Team
The defining scaling lab of the 2020s.
Applications
- Advanced Reasoning
- Vision-Language parsing
- Real-time Voice AIs
Key Papers
- GPT-4 Technical Report
OpenAI · arXiv · 2023