GPT-4 & Large Language Models
The spark of artificial general reasoning.
Explore this event on the interactive timeline →OpenAI released GPT-4, a massive multimodal Large Language Model (LLM) trained on a significant portion of the public internet. It demonstrated unprecedented reasoning, coding, and language comprehension abilities.
Key Numbers
- Parameters (2012 AlexNet)
- 60 Million
- Parameters (2023 GPT-4)
- ~1.76 Trillion
Verified Facts
- GPT-4 reportedly utilized a "Mixture of Experts" architecture with roughly 1.76 Trillion parameters—the simulated synapses of its artificial brain.
- It was capable of passing the Uniform Bar Exam in the 90th percentile, writing complex software from scratch, and reasoning through highly complex, novel logic puzzles.
- This release triggered a massive global arms race, with tech monoliths (Google, Anthropic, Meta) mobilizing tens of billions of dollars to build massive GPU data centers to train the next generation of models.