Benchmark notes
First OpenAI model designed natively for text, audio, and image in a single model ('omni'). Retired from ChatGPT on February 13, 2026. Still available in the API as of July 2026; voice mode continues to use GPT-4o or GPT-4o mini depending on plan. The image generation model GPT Image 1 (gpt-image-1) was based on GPT-4o architecture.