AI capability timeline

Range means the collection of practical capability types. Performance means quality or speed within them. Access and connections make capabilities easier to employ or combine; they do not automatically add a new capability.

The starting bundle already includes writing, reasoning, coding assistance, image generation, speech, audio processing, and emerging music generation. These existed before ChatGPT launched. The timeline tracks their subsequent expansion and integration, rather than claiming that each product invented its category.

Lite selects 24 milestones that opened a substantial new kind of work, combined capabilities into a meaningful new workflow, or delivered a major quality, speed, or deployment improvement. Routine model updates and repeated access expansions stay in Full. This is an editorial selection, not a benchmark ranking.

Provider and milestone-type filters apply to the list, playback, navigation, and slider. The graph continues to show the overall bundle at each selected date.

The rectangle is a conceptual illustration. Width reflects the capability categories represented in this story; height provides a visual progression through performance milestones. Neither is a validated index, and a competing model release does not necessarily advance the global frontier. A real comparison requires capability-specific benchmarks with comparable testing conditions.

Dates refer to first documented access or the beginning of a rollout. Some releases are limited previews or restricted by plan, country, or invitation. Month-only dates retain that precision. Examples describe supported uses, not a guarantee that no earlier system could perform them. Sources and limitations accompany each milestone. Research snapshot: September 30, 2026.