OpenAI API End of Life (EOL) Dates & Support Timeline
Complete end-of-life dates, support windows, and security status for all OpenAI API versions. Data from endoflife.date, checked against the vendor's own lifecycle pages where we have read them (share on /accuracy). Updated at every deploy.
OpenAI API gpt-5-pro is supported until December 11, 2026, its end-of-support date. The next OpenAI API version to reach end of life is o4-mini, on October 23, 2026. 9 of 29 tracked OpenAI API versions are past end of life; the most recent to reach it, gpt-3.5-turbo-instruct, did so on September 28, 2026. The full OpenAI API lifecycle, with every version's release and end-of-support date, is in the table below.
| Version | Latest Release | Release Date | EOL Date | Days | Status |
|---|---|---|---|---|---|
| text-davinci-002 | text-davinci-002 | Mar 15, 2022 | Jan 4, 2024 | 1013 days past EOL | EOL |
| text-davinci-003 | text-davinci-003 | Nov 28, 2022 | Jan 4, 2024 | 1013 days past EOL | EOL |
| gpt-3.5-turbo-0301 | gpt-3.5-turbo-0301 | Mar 1, 2023 | Jun 13, 2023 | 1218 days past EOL | EOL |
| gpt-4-0314 | gpt-4-0314 | Mar 14, 2023 | Jun 13, 2023 | 1218 days past EOL | EOL |
| gpt-3.5-turbo-0613 | gpt-3.5-turbo-0613 | Jun 13, 2023 | Sep 13, 2024 | 760 days past EOL | EOL |
| gpt-4-0613 | gpt-4-0613 | Jun 13, 2023 | Oct 23, 2026 | 10 days remaining | Warning |
| gpt-3.5-turbo-instruct | gpt-3.5-turbo-instruct | Sep 18, 2023 | Sep 28, 2026 | 15 days past EOL | EOL |
| gpt-3.5-turbo-0125 | gpt-3.5-turbo-0125 | Jan 25, 2024 | Oct 23, 2026 | 10 days remaining | Warning |
| gpt-4-turbo | gpt-4-turbo-2024-04-09 | Apr 9, 2024 | Oct 23, 2026 | 10 days remaining | Warning |
| gpt-4o-2024-05-13 | gpt-4o-2024-05-13 | May 13, 2024 | Oct 23, 2026 | 10 days remaining | Warning |
| codex-mini-latest | codex-mini-latest | May 15, 2025 | Feb 12, 2026 | 243 days past EOL | EOL |
| gpt-4o-mini | gpt-4o-mini-2024-07-18 | Jul 18, 2024 | TBD | Supported | Active |
| gpt-4o-2024-08-06 | gpt-4o-2024-08-06 | Aug 6, 2024 | TBD | Supported | Active |
| o1-mini | o1-mini | Sep 12, 2024 | Oct 27, 2025 | 351 days past EOL | EOL |
| o1-preview | o1-preview | Sep 12, 2024 | Oct 28, 2025 | 350 days past EOL | EOL |
| gpt-4o-2024-11-20 | gpt-4o-2024-11-20 | Nov 20, 2024 | TBD | Supported | Active |
| o1 | o1-2024-12-17 | Dec 5, 2024 | Oct 23, 2026 | 10 days remaining | Warning |
| o3-mini | o3-mini-2025-01-31 | Jan 31, 2025 | Oct 23, 2026 | 10 days remaining | Warning |
| o1-pro | o1-pro-2025-03-19 | Mar 19, 2025 | Oct 23, 2026 | 10 days remaining | Warning |
| gpt-4.1 | gpt-4.1 | Apr 14, 2025 | TBD | Supported | Active |
| gpt-4.1-mini | gpt-4.1-mini | Apr 14, 2025 | TBD | Supported | Active |
| gpt-4.1-nano | gpt-4.1-nano | Apr 14, 2025 | Oct 23, 2026 | 10 days remaining | Warning |
| o3 | o3 | Apr 16, 2025 | Dec 11, 2026 | 59 days remaining | Warning |
| o4-mini | o4-mini | Apr 16, 2025 | Oct 23, 2026 | 10 days remaining | Warning |
| o3-pro | o3-pro-2025-06-10 | Jun 10, 2025 | Dec 11, 2026 | 59 days remaining | Warning |
| gpt-5-nano | gpt-5-nano-2025-08-07 | Aug 7, 2025 | Dec 11, 2026 | 59 days remaining | Warning |
| gpt-5-mini | gpt-5-mini-2025-08-07 | Aug 7, 2025 | Dec 11, 2026 | 59 days remaining | Warning |
| gpt-5 | gpt-5-2025-08-07 | Aug 7, 2025 | Dec 11, 2026 | 59 days remaining | Warning |
| gpt-5-pro | gpt-5-pro-2025-10-06 | Oct 6, 2025 | Dec 11, 2026 | 59 days remaining | Warning |
Live EOL badge for your README — updates automatically at every build.
[](https://endoflife.ai/openai-api)
OpenAI API release and support pattern
Of the 29 release lines tracked, 9 are past end of life and 20 are still within support. The newest line, gpt-5-pro, was released October 6, 2025. The next scheduled end of life is o4-mini on October 23, 2026 — 9 days from this build. The longest-supported line currently published is o3, to December 11, 2026.
Computed from the tracked release data at every build — the figures above are the dates in the table, restated, not an editorial estimate.
What does OpenAI API end of life mean for your organization?
When a version of OpenAI API reaches end of life, the maintainers stop issuing security patches. Vulnerabilities discovered after this date are publicly disclosed on the National Vulnerability Database, exploit code appears on GitHub, and your systems remain permanently exposed.
The CVE blind spot: A CVE-based scan does not show the accumulation of unpatched vulnerabilities in EOL software. Some scanners flag unsupported versions of common products, but coverage varies, and none of them give you the date in advance. With a zero-day, nobody knows about the vulnerability. With EOL software, the vulnerability is public — listed, rated, and often weaponized — but no vendor patch is coming. This is one of the most dangerous gaps in enterprise security posture.
Organizations running EOL OpenAI API should treat it as a vulnerability class in their risk register, apply compensating controls (network segmentation, enhanced monitoring, access restriction), and prioritize migration to a supported version.
Your options when OpenAI API reaches end of life
There are four realistic paths. Which one is right depends far more on how long the system needs to keep running than on the version you are on today.
| Option | Best when | Watch out for |
|---|---|---|
| Move to a supported version | The system will outlive the current version's support window and you can schedule the work | Budget for compatibility testing, not just the upgrade itself — that is where the real cost sits |
| Replace the component | The project is unmaintained, or a better-supported alternative exists | The replacement has its own lifecycle and its own end-of-life date — you are changing vendors, not escaping the problem |
| Commercial extended support | Not available for OpenAI API from the original maintainer | Third-party vendors patch some end-of-life software on a custom basis; availability depends on the specific version and source access |
| Document accepted risk | The system is isolated, short-lived, or scheduled for retirement | Only defensible with real compensating controls and a written decision — auditors treat undocumented end-of-life software as a finding |
How to choose, in one question: how many more years does this system need to run? Under a year, extended support (where it exists) is usually cheaper than an emergency migration. One to three years, migrate — support purchased repeatedly costs more than doing the project once, and the project only gets harder as the version ages. Indefinitely, migrate now and plan the next migration before it surprises you.
Worth saying plainly, because vendors selling extended support rarely will: extended support is often the more expensive choice over a multi-year horizon. It is the right answer when a deadline is genuinely unreachable, not as a default. Compare the annual fee against the one-time cost of migrating before committing to either.
Upload requirements.txt, package.json, or Gemfile — full EOL report instantly.
Open Stack Scanner →