OpenAI sets a 23 October date to remove GPT-4, GPT-3.5 Turbo, o1, o3-mini and a batch of older API models

One from discovery tonight, and a big one: OpenAI's API deprecations page now sets a single 23 October 2026 shutdown for a batch of older chat and reasoning models — gpt-4 (gpt-4-0613), gpt-3.5-turbo, gpt-4-turbo, gpt-4.1-nano, the gpt-4o-2024-05-13 snapshot, o1, o1-pro, o3-mini and o4-mini, plus fine-tunes built on them — moving developers to the gpt-5.6 family. Anything that pins one of those ids by string breaks that day; the surviving gpt-4o, gpt-4.1 and gpt-4.5 are not on the list. On the board: OpenSSL 3.0 reaches end-of-life in two days (7 September), and OpenAI's Sora 2 video API shuts down on 24 September.

One item tonight, and it is a substantial one, surfaced through a discovery search and then written from OpenAI’s own API deprecations page. No corrections are outstanding. Tonight’s three re-verification flags were again the Microsoft pages that render their dates in JavaScript (Sentinel SAP connector, Azure Automation Python/PowerShell runtimes, Exchange Web Services) — each re-confirmed correct by hand against Microsoft’s own release-communications API and page metadata, not data errors.

New: OpenAI removes GPT-4, GPT-3.5 Turbo, o1, o3-mini and more from the API — 23 October 2026

OpenAI’s deprecations page now sets a single shutdown date of 23 October 2026 for a large batch of older chat and reasoning models. After that date, a request that names one of these ids returns an error instead of a completion:

gpt-3.5-turbo (gpt-3.5-turbo-0125)
gpt-4 (gpt-4-0613), gpt-4-1106-preview
gpt-4-turbo (gpt-4-turbo-2024-04-09)
gpt-4.1-nano, the gpt-4o-2024-05-13 snapshot, gpt-image-1
o1, o1-pro, o3-mini, o4-mini
fine-tuned models built on gpt-3.5-turbo, gpt-4, gpt-4.1-nano, babbage-002 and davinci-002

OpenAI points developers to its gpt-5.6 family (gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna), with the exact per-model mapping on the deprecations page. Read the version numbers carefully: the general gpt-4o, gpt-4.1 and gpt-4.5 models are not on the list — only the specific ids above. The retiring gpt-4.1-nano is the nano tier, not gpt-4.1; and it is the dated gpt-4o-2024-05-13 snapshot that goes, not gpt-4o.

For anyone who uses the current default this is a non-event. It bites where a model id is hardcoded and forgotten — a wrapper that defaults to gpt-4, an env var set two years ago, a batch job pinned to gpt-3.5-turbo, an eval config still naming o1. Because a swapped model can change output shape, latency and cost, this is a change to test, not a blind find-and-replace — re-run your evals against the gpt-5.6 replacement before the date. This entry is detectable: the lastcall CLI fingerprints the retiring OpenAI-exclusive ids (never the surviving gpt-4o/gpt-4.1/gpt-4.5), so a pinned model that would break in October surfaces as a warning in your terminal now. It is one wave of a rolling schedule — the same page lists the GPT-5 snapshots, o3 and o3-pro for 11 December, alongside the image (1 December) and audio/realtime (into 2027) models already on our board.

On the board

7 September — two days out: OpenSSL 3.0 reaches end-of-life.
4 September — still watched: Google’s reported removal of Google Assistant on mobile still has no first-party dated page (only “over the coming months”), so the entry stays signaled with no countdown until Google publishes a date.

Further out: the GitHub Copilot MAI-Code-1-Flash deprecation (10 September), the Relay app and Microsoft Sentinel SAP connector agent retirements (14 September), the Azure Maps Render v1 API retirement (17 September), OpenAI’s Sora 2 video API (24 September), a cluster on 30 September and 1 October, the GitHub Copilot model deprecations (2 October), then OpenAI’s own 23 October removals above, .NET 8 and 9 end of support (10 November) and PostgreSQL 14 end-of-life (12 November). The full board with live countdowns is on the calendar.