AI Model Retirement Lint

Three of the model IDs in a typical llm.config.js already return an error.

A retired model ID is not a syntax error. It is a valid string, it type-checks, and every test goes green - right up to the day the vendor's API answers 404 instead of a completion.

// llm.config.js - committed by an AI coding assistant
export const MODELS = {
  chat:      "gpt-4-turbo",                    // shuts down 2026-10-23
  cheap:     "gpt-3.5-turbo-0125",             // shuts down 2026-10-23
  reasoning: "o3-mini-2025-01-31",             // shuts down 2026-10-23
  claude:    "claude-3-5-sonnet-20241022",     // RETIRED 2025-10-28
  fast:      "claude-3-haiku-20240307",        // RETIRED 2026-04-20
  gemini:    "gemini-2.5-flash",               // shuts down 2026-10-16
  voice:     "whisper-1",                      // RETIRED 2026-08-26
  images:    "gpt-image-1",                    // shuts down 2026-10-23
};

Eight findings in that file. Three of them fail today.

Why your assistant will not catch this

The model that autocompleted claude-3-5-sonnet-20241022 into your config was trained before Anthropic announced its retirement. Ask it, and it confirms the dead ID as correct. Its own model ID is often on the same list.

What is in the box

• 33 rules across OpenAI, Anthropic and Gemini, each carrying the vendor's published date and the vendor's own recommended replacement.
• Three severities: error = the call fails now, warn = a published shutdown date ahead, info = the vendor has named an earliest retirement date.
• Covers the 2026-10-23 OpenAI sweep (12 model IDs), the 2026-09-28 and 2026-10-16 shutdowns, the 2026-12-11 GPT-5 snapshot cut, and every Claude model retired to date.

Free, with no key

• Check this file for retired model IDs - all 33 rules, line numbers, dates, replacements
• List every model ID and date I know - the whole dated table, browsable
• Show the last report

Nothing is hidden, timed, counted, or watermarked. Checking the file you have open is free and complete.

With a licence key ($29 once)

A different job, not a bigger portion of the same one:

Get the full version - $29

One key per person or team seat. 7-day full refund.

Sources

Every date ships from the vendor's own deprecation page, read on 2026-09-09: OpenAI API deprecations, Anthropic model deprecations, Google Vertex AI / Gemini model retirements.

Install

ext install ai-model-retirement-lint