On May 13, 2024, OpenAI unveiled GPT-4o (the "o" stands for "omni")—their first truly multimodal model that natively processes text, voice, and vision together.

Not separate models stitched together. One unified model understanding all three simultaneously.

What Made It Different

True multimodality: Single model, not separate voice/vision/text models connected Real-time voice: Natural conversation with minimal latency Emotion detection: Understood tone, inflection, emotional context Vision integration: Analyzed images while talking about them Free for all: GPT-4o became free tier, not Plus-exclusive

The Demos

OpenAI's launch demos were stunning:

• Real-time tutoring with voice and visual math problems
• Translating between speakers in different languages
• Analyzing code on screen while discussing it
• Singing and emotional voice responses

It felt like AI from science fiction.

The Speed

GPT-4o was 2x faster than GPT-4 while being more capable. This made real-time voice conversation actually work—no awkward pauses.

The Accessibility

Most importantly: GPT-4o became the free tier for ChatGPT. Everyone could access frontier AI, not just $20/month subscribers.

This democratized access dramatically.

Where Are They Now?

GPT-4o remains OpenAI's standard model for most users. The voice mode particularly impressed users as genuinely conversational AI.

May 13, 2024 was when AI assistants started feeling less like chatbots and more like actual assistants—seeing, hearing, and speaking naturally.