GPT-4o and GPT-4.1 leave the OpenAI API on Jan 14
Hardcoded gpt-4o, gpt-4o-mini, gpt-4.1 and gpt-4.1-mini calls stop in 92 days. OpenAI says move to gpt-6-luna and start at reasoning none.
If any app, script or automation you run calls gpt-4o, gpt-4o-mini, gpt-4.1 or gpt-4.1-mini through the OpenAI API, put January 14, 2027 on your calendar and start testing gpt-6-luna this week. OpenAI's own migration note says to begin with reasoning effort set to none, which it expects to cover most GPT-4.x use cases.
What changed#
On Oct 14, OpenAI added a "GPT-4 series model deprecations" entry to its API deprecations page. It says developers using these models were notified that they will be removed from the API on January 14, 2027:
| Retiring snapshot | Replacement |
|---|---|
gpt-4o-2024-08-06, gpt-4o-2024-11-20 | gpt-6-luna |
gpt-4o-mini-2024-07-18 | gpt-6-luna |
gpt-4.1-2025-04-14 | gpt-6-luna |
gpt-4.1-mini-2025-04-14 | gpt-6-luna |
Fine-tuned versions follow the same shutdown schedule as their base models.
Before: these were open-ended GA models. After: a hard date 92 days out. The same page says generally available models get "at least 6 months" of notice unless safety or compliance concerns require faster. The Oct 14 entry gives no reason for the shorter window.
Why it matters#
These model names sit in a lot of old glue: support bots, spreadsheet scripts, Zap-style automations and internal tools nobody has touched since 2024. The bare alias matters too. OpenAI's GPT-4o page lists gpt-4o-2024-08-06 as its default snapshot, and the 4.1, 4.1 mini and 4o mini pages likewise point to the retiring snapshots. The notice does not say what the bare names do after Jan 14, so plan as if they stop.
On list price the move is a cut. Our math for a job that sends 10M input and gets 2M output tokens a month, no caching:
| Model | Input / Output per 1M | Monthly |
|---|---|---|
| GPT-4o | $2.50 / $10 | $45.00 |
| GPT-4.1 | $2 / $8 | $36.00 |
| GPT-4.1 mini | $0.40 / $1.60 | $7.20 |
| GPT-4o mini | $0.15 / $0.60 | $2.70 |
| GPT-6 Luna | $0.10 / $0.50 | $2.00 |
The catch: Luna is a reasoning model that defaults to medium effort, and reasoning tokens are billed as output. If reasoning doubled that job's output to 4M tokens, Luna would cost $3.00, more than GPT-4o mini today.
How to use it#
- Search everything, not just code:
rg -n "gpt-4o|gpt-4\.1|o4-mini|o3-mini|gpt-4-turbo" .plus your no-code tools' model dropdowns and env files. - Ask the API which models have a date. OpenAI's deprecations page says
GET /v1/modelsreturns an optionalshutdown_date:
curl -s https://api.openai.com/v1/models -H "Authorization: Bearer $OPENAI_API_KEY" \
| jq -r '.data[] | select(.shutdown_date != null) | "\(.id)\t\(.shutdown_date)"'
- Follow OpenAI's four-step plan: record a baseline (quality, latency, cost) on real requests, test Luna at
none, tune prompts and effort, then roll out to a slice of traffic first. - Swap the model and pin effort explicitly, for example
"model": "gpt-6-luna", "reasoning_effort": "none"in Chat Completions. - If you raise effort above
none, removetemperature,top_pandtop_logprobs(andlogprobsin Chat Completions), per OpenAI's GPT-6 guide.
Copyable prompt for Codex or ChatGPT Work:
Find every OpenAI model ID in this repo and its configs. For each hit list file:line, the model, the endpoint (Chat Completions or Responses), and whether it sends temperature, top_p, logprobs or tools. Flag anything in OpenAI's Oct 23, 2026 or Jan 14, 2027 shutdown lists. Then propose a gpt-6-luna patch that sets reasoning_effort to none, without changing behavior, and list what I must eval before merging.
Good to know#
- Nearer deadline: a separate list shuts down Oct 23, 2026, nine days from now, including
gpt-4o-2024-05-13,gpt-4.1-nano,gpt-4-turbo,o3-mini,o4-mini,o1andgpt-image-1. - Tools in Chat Completions: Luna supports function calling there only with
reasoning_effortset tonone. For reasoning plus tools, use the Responses API. - Fine-tunes have no like-for-like home. Luna's model page marks fine-tuning "Not supported," and active customers lose new fine-tuning jobs on Jan 6, 2027.
- Long prompts: Luna prices requests over 272K input tokens at 2x input and 1.5x output for the whole request. Luna's context is 1,050,000 tokens versus 128,000 on GPT-4o.
- Watch for OpenAI's notification email, a published quality comparison (none exists yet), and
shutdown_dateappearing on any other model you use.
Sources#
- OpenAI API deprecations: https://developers.openai.com/api/docs/deprecations (checked 2026-10-14 5:57 AM ET)
- OpenAI Help Center, Migrating from GPT-4.x to GPT-6 Luna: https://help.openai.com/en/articles/20001600-migrating-from-gpt-4x-to-gpt-6-luna (checked 5:58 AM ET)
- GPT-6 Luna model page: https://developers.openai.com/api/docs/models/gpt-6-luna (checked 5:58 AM ET)
- Using GPT-6 guide: https://developers.openai.com/api/docs/guides/latest-model (checked 5:59 AM ET)
- Reasoning models guide: https://developers.openai.com/api/docs/guides/reasoning (checked 5:59 AM ET)
- GPT-4o, GPT-4.1, GPT-4o mini, GPT-4.1 mini model pages: https://developers.openai.com/api/docs/models/gpt-4o , https://developers.openai.com/api/docs/models/gpt-4.1 , https://developers.openai.com/api/docs/models/gpt-4o-mini , https://developers.openai.com/api/docs/models/gpt-4.1-mini (checked 5:58 AM ET)
Last checked: 2026-10-14
Steps
- Find every gpt-4o, gpt-4o-mini, gpt-4.1 and gpt-4.1-mini string in code, configs and no-code tools.
- List models with a shutdown_date via GET /v1/models.
- Record a baseline on real requests, then test gpt-6-luna at reasoning none.
- Remove temperature, top_p, top_logprobs and logprobs if you raise reasoning effort above none.
- Roll out on a slice of traffic, watch cost and quality, then switch.