Eleven model snapshots retire on October 23. Find them in your repo before your users do
OpenAI shuts down eleven model snapshots on 2026-10-23, among them gpt-4-turbo , gpt-4o-2024-05-13 , o3-mini and o4-mini . The GPT-5 and o3 snapshots follow on 2026-12-11. Anthropic and Google retire models on their own schedules too. Most codebases find out when requests start returning 404. Finding what retires I wrote modelshift to answer two questions from the terminal. Which model IDs in my…
On 2026-10-23, OpenAI ceased eleven model snapshots, including gpt-4-turbo, gpt-4o-2024-05-13, o3-mini and o4-mini. Subsequent retirements include GPT-5 and o3 snapshots on 2026-12-11. Other companies, such as Anthropic and Google, also end their models according to their own timelines. Codebases typically discover model retirements when requests begin returning 404 errors.
To address this issue, a reporter developed a tool called modelshift, which can be used in a terminal to identify retiring model IDs and assess whether their replacements maintain similar functionality. The modelshift scan command analyzes a repository, listing each model ID along with its file and line, provider, lifecycle status, and a recommended replacement.
The status is derived from a bundled registry generated from the providers' deprecation pages. Modelshift can also function as a CI check, as its exit code is 1 when a model retires within the specified timeframe. The reporter ran modelshift across four unrelated open-source repositories, uncovering 131 distinct retiring model IDs referenced 1,649 times across 296 files.
With the modelshift fix command, the reporter can generate a diff that swaps each ID for its replacement and eliminates any parameters that the new model rejects. When swapping model strings, the reporter must also ensure the new model still returns valid JSON, invokes tools with identical arguments, and maintains comparable performance.
The replay command compares prompts against both the old and new models, providing an HTML report detailing differences in JSON validity, refusal rates, tool calls, output length, latency, and text similarity. Modelshift requires Node 20 or later and is not yet available on npm; users must run it using npx from the GitHub repository.
The bundled registry is manually maintained from provider pages, so incorrect dates are the most probable issue. If a reporter discovers a discrepancy, they should open an issue on GitHub. The modelshift code is open-source under the MIT license and available at https://github.com/Arthur031221/modelshift.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
