# How does ChatGPT separate current and legacy models in its model picker?

Source: Lazyweb Research
Author: Ali Abouelatta, Lazyweb Research
Published: 2026-09-09
Reviewed: September 2026
Category: [AI interface decisions](https://design.lazyweb.com/blog/category/ai-interface-decisions)
Tags: ai, navigation, progressive-disclosure
HTML: https://design.lazyweb.com/blog/chatgpt-model-picker-progressive-disclosure
Markdown: https://design.lazyweb.com/blog/chatgpt-model-picker-progressive-disclosure.md

**Answer.** In Lazyweb’s January 2026 capture, the first menu presents four GPT-5.2 choices with short descriptions. A separate Legacy models row opens a longer, scrollable list. The pattern keeps the primary decision compact while retaining access to older options. This is a historical UI example, not a list of models available today. [1]

## What the two screenshots show

The first screen groups Auto, Instant, Thinking and Pro beneath the GPT-5.2 heading. Auto carries a checkmark. Each option pairs a short name with an explanation of its intended behavior. A divider separates Legacy models from these primary choices. [1]

The second screenshot shows the expanded legacy list, including several model generations and a visible scrollbar. The conversation composer remains visible behind the menu. These are two captured menu states, not a complete recording of selection and response behavior. [1]

## The design decision

There are two different jobs: choose a behavior for the next response, or deliberately return to an older model. The captured hierarchy gives those jobs separate levels. Our interpretation is that this can reduce the number of equally prominent choices; the screenshots do not measure whether people choose faster. [1]

## When to use the pattern

Consider a short primary list when a product has a few common modes and many specialist or older options. Give each primary choice a meaningful description, expose selection state, and label the secondary collection clearly. Do not copy a historical model roster into a current product. This is design guidance, not an observed experiment result.

## Methodology

Two public screenshots from one historical ChatGPT mobile flow, inspected individually. No current model-availability, interaction-timing or conversion claim is made.

## Visual references

Real, in-market app screens tracked by Lazyweb Research:

- ChatGPT: https://zlfyzdmohcskkucuunmk.supabase.co/storage/v1/object/public/screenshots/prod11_chatgpt/backlog/2026_01_13_10-34-32_02C5AA88.png
- ChatGPT: https://zlfyzdmohcskkucuunmk.supabase.co/storage/v1/object/public/screenshots/prod11_chatgpt/backlog/2026_01_13_10-34-34_117A3DFA.png

## Sources & citations

- [1] ChatGPT model-picker capture. Two public screenshots; page metadata updated January 2026. Visually inspected September 9, 2026. [Open source](https://design.lazyweb.com/flow/chatgpt/model-picker)

## Related questions

- [What comes between sign-in and the trial reminder in Cal AI’s paywall flow?](https://design.lazyweb.com/blog/cal-ai-sign-in-trial-reminder-sequence)

## Explore the evidence

- [ChatGPT model picker](https://design.lazyweb.com/flow/chatgpt/model-picker) — Two captured screens · January 2026
- [Claude feature-access prompt](https://design.lazyweb.com/flow/claude/locked-feature-paywall)

## Continue your research

[Connect Lazyweb to your coding agent](https://www.lazyweb.com/mcp-install)