Moving code off gpt-4o, gpt-4o-mini and Claude Sonnet 4.5

The model `computer-use-preview-2025-03-11` has been deprecated

Short answer: Search your code, environment files and config for the retiring model IDs and change them before each date: gpt-4o, gpt-4o-mini, gpt-4.1 and gpt-4.1-mini snapshots stop on 14 January 2027 (replacement gpt-6-luna), claude-sonnet-4-5-20250929 on 30 November 2026. gpt-6-luna is a reasoning model, so check tool calls and sampling settings, not only the ID.

Scan your code and config for retiring models (free, in your browser)

Where the model is set

Model IDs hide in more places than the client call: .env files, YAML or TOML config, LangChain constructors (ChatOpenAI(model=…)), OpenRouter IDs (openai/gpt-4o-mini), Bedrock IDs (anthropic.claude-sonnet-4-5-20250929-v1:0), Vertex IDs (claude-sonnet-4-5@20250929) and fine-tune IDs (ft:gpt-4o-mini-2024-07-18:…). The scanner recognises all of these forms. The error above is what Chat Completions returned the day after OpenAI's July 2026 shutdown; the Responses API answered "Model not found …", and an ID OpenAI does not recognise returns 404 model_not_found. OpenAI does not document the exact text.

Fix

Search the repository: git grep -nE "gpt-4o|gpt-4\.1|gpt-5(-mini|-nano)?\b|\bo3\b|whisper-1|tts-1|claude-sonnet-4-5", or paste files into the scanner.
Replace each ID with the replacement in the scanner's list. Keep the ID in one config value so the next change is one line.
gpt-6-luna on Chat Completions with tools: set reasoning_effort="none", or move the call to the Responses API. With any other effort, remove temperature, top_p, top_logprobs and logprobs. Use max_completion_tokens instead of max_tokens.
Claude: move off claude-sonnet-4-5-20250929 to claude-sonnet-5-5 before 30 November 2026. If you upgrade to Python anthropic 2.0.0 or @anthropic-ai/sdk 1.0.0, stop sending temperature, top_p and top_k: API version 2026-10-21 rejects them with a 400.
Fine-tunes: a fine-tune stops with its base snapshot. Retrain on a current base before 6 January 2027, when new fine-tuning jobs end.
Add a check in CI: compare the IDs in your code with /model-registry.json or with the shutdown_date field of OpenAI's List models endpoint.

Many workflows or services to move? The AI Model Shutdown Kit patches every n8n workflow through the API with a backup of each, scans code and config, and compares the old and new model on your own prompts (answers, tool calls, tokens, cost) before you switch. 18-page guide included.

Get the kit ($29, agencies $79)

Watch out

Bare aliases such as gpt-4o-mini are not named in OpenAI's 14 January notice, but they point only to snapshots that are. Do not rely on the alias moving to a new model.
Test with a day of real prompts: a reasoning model can answer differently and use more output tokens.

Sources

Other platforms

AI model shutdown calendar and scanner

Facts checked 28 October 2026. Not affiliated with OpenAI, Anthropic, Google or Python, Node.js and LangChain code.

Did this fix it?