Skip to content
RRanked By Task

This Week

OpenAI's listed shutdown date arrives for older API snapshots

OpenAI's deprecations page, from a 22 April 2026 notice, sets 23 October 2026 as the shutdown date for a set of older GPT, o-series, and image snapshots.

Official source (opens in a new tab)

OpenAI's deprecations page, under a notice dated 22 April 2026, says access to a set of older models will be shut down on the dates in the table. For the rows dated 23 October 2026, the substitute in the last cell is a newer model. Where a row also prints other IDs between the snapshot and that substitute, this note lists those IDs as they appear on the row.

The page pairs gpt-3.5-turbo-0125 with gpt-3.5-turbo and gpt-3.5-turbo-completions, and names gpt-5.6-terra as the substitute. It pairs gpt-4-0613 with gpt-4, gpt-4-0613-completions, and gpt-4-completions, substitute gpt-5.6-sol. gpt-4-1106-preview substitutes gpt-5.6-sol. gpt-4-turbo is listed with gpt-4-turbo-2024-04-09 and gpt-4-turbo-completions, substitute gpt-5.6-sol. gpt-4.1-nano is listed with gpt-4.1-nano-2025-04-14, substitute gpt-5.6-luna. gpt-4o-2024-05-13 substitutes gpt-5.6-sol. gpt-image-1 substitutes gpt-image-2.5-sunburst or gpt-image-2.5-flare. o1-2024-12-17 is listed with o1, and o1-pro-2025-03-19 with o1-pro, both substituting gpt-5.6-sol, the pro row with reasoning.mode set to pro. o3-mini-2025-01-31 is listed with o3-mini, substitute gpt-5.6-sol. o4-mini-2025-04-16 is listed with o4-mini, substitute gpt-5.6-terra. A fine-tune of o4-mini-2025-04-16 substitutes gpt-5.6-terra.

A second table on the same date names fine-tuned replacements: ft-gpt-3.5-turbo and ft-babbage-002 and ft-davinci-002 to gpt-5.6-terra, ft-gpt-4 to gpt-5.6-sol, and ft-gpt-4.1-nano-2025-04-14 to gpt-5.6-luna. The page's wording is that access will be shut down on that date. This post does not claim a live request was refused.

Who it's for: anyone still calling these snapshot IDs from the API.

Fits: Best for Writing stays Claude, and ChatGPT stays the generalist chat product on that list. These rows are old API snapshots, not the current chat model. The verdicts are unchanged.