OpenAI Retires Older GPT Models (Feb 2026)
OpenAI Retires Older GPT Models (Feb 2026)
OpenAI
29 janv. 2026


Pas sûr de quoi faire ensuite avec l'IA?
Évaluez la préparation, les risques et les priorités en moins d'une heure.
Pas sûr de quoi faire ensuite avec l'IA?
Évaluez la préparation, les risques et les priorités en moins d'une heure.
➔ Téléchargez notre kit de préparation à l'IA gratuit
OpenAI will retire GPT‑4o, GPT‑4.1, GPT‑4.1 mini and OpenAI o4‑mini in ChatGPT on 13 February 2026, alongside the previously announced retirement of GPT‑5 (Instant and Thinking). The aim is a faster, more reliable experience while focusing resources on current models. No API changes at this time.
What’s Changing & Why
Which models are leaving? GPT‑4o, GPT‑4.1, GPT‑4.1 mini and OpenAI o4‑mini leave ChatGPT on 13 Feb 2026. GPT‑5 (Instant & Thinking) retirement was announced earlier.
Why retire models? Consolidation improves reliability, performance and support, allowing investment in the latest models and features.
API status: The retirement applies to ChatGPT; no API changes for now. If you depend on legacy model snapshots in the API, monitor the deprecations page and Azure notices.
Practical Guidance (Users & Admins)
Individuals (ChatGPT):
You don’t need to take action. Your conversations will continue to work; newer default models will be used.
If you liked a legacy model’s tone, update Custom Instructions to guide style and behaviour.
Workspace admins:
Communicate the change and suggest updated templates/prompts.
Review any team playbooks referencing specific legacy model names.
For regulated teams, update screenshots/process docs showing the model switcher.
Developers:
For API apps, check the Deprecations page periodically. If using Azure OpenAI, note separate retirement timelines and auto‑upgrade plans.
Example Transitions
From GPT‑4o → GPT‑5.2: Re‑test saved prompts; add tone instructions (e.g., concise, warm, British English).
From GPT‑4.1 mini → current lightweight model: Validate latency/quality; adjust temperature and max tokens if needed.
From o4‑mini → current “o‑class” replacement: Confirm tool‑use behaviours and output formats.
Risks & Good Practice
Expectation drift: Legacy models may have had distinct “feel.” Capture style via instructions/examples.
Evaluation: Re‑run quick acceptance tests on key prompts (accuracy, format, latency).
Change control: Document model settings; pin versions where supported; keep rollback notes for prompts.
FAQs
What models are being retired?
GPT‑4o, GPT‑4.1, GPT‑4.1 mini and OpenAI o4‑mini leave ChatGPT on 13 February 2026. GPT‑5 (Instant/Thinking) retirement was previously announced.
Why is OpenAI retiring these models?
To streamline the experience, improve performance and reliability, and focus resources on newer models.
Will there be changes to the API?
No changes at this time for the API; continue to monitor official deprecation notices for snapshots and Azure timelines.
How will this affect users?
Most users won’t need to act. The default model selection will shift to current versions, with improved performance and stability.
How can I preserve a legacy model’s tone?
Use Custom Instructions and a short style block (examples of tone, phrasing, and do/don’t rules) to approximate previous behaviour.
OpenAI will retire GPT‑4o, GPT‑4.1, GPT‑4.1 mini and OpenAI o4‑mini in ChatGPT on 13 February 2026, alongside the previously announced retirement of GPT‑5 (Instant and Thinking). The aim is a faster, more reliable experience while focusing resources on current models. No API changes at this time.
What’s Changing & Why
Which models are leaving? GPT‑4o, GPT‑4.1, GPT‑4.1 mini and OpenAI o4‑mini leave ChatGPT on 13 Feb 2026. GPT‑5 (Instant & Thinking) retirement was announced earlier.
Why retire models? Consolidation improves reliability, performance and support, allowing investment in the latest models and features.
API status: The retirement applies to ChatGPT; no API changes for now. If you depend on legacy model snapshots in the API, monitor the deprecations page and Azure notices.
Practical Guidance (Users & Admins)
Individuals (ChatGPT):
You don’t need to take action. Your conversations will continue to work; newer default models will be used.
If you liked a legacy model’s tone, update Custom Instructions to guide style and behaviour.
Workspace admins:
Communicate the change and suggest updated templates/prompts.
Review any team playbooks referencing specific legacy model names.
For regulated teams, update screenshots/process docs showing the model switcher.
Developers:
For API apps, check the Deprecations page periodically. If using Azure OpenAI, note separate retirement timelines and auto‑upgrade plans.
Example Transitions
From GPT‑4o → GPT‑5.2: Re‑test saved prompts; add tone instructions (e.g., concise, warm, British English).
From GPT‑4.1 mini → current lightweight model: Validate latency/quality; adjust temperature and max tokens if needed.
From o4‑mini → current “o‑class” replacement: Confirm tool‑use behaviours and output formats.
Risks & Good Practice
Expectation drift: Legacy models may have had distinct “feel.” Capture style via instructions/examples.
Evaluation: Re‑run quick acceptance tests on key prompts (accuracy, format, latency).
Change control: Document model settings; pin versions where supported; keep rollback notes for prompts.
FAQs
What models are being retired?
GPT‑4o, GPT‑4.1, GPT‑4.1 mini and OpenAI o4‑mini leave ChatGPT on 13 February 2026. GPT‑5 (Instant/Thinking) retirement was previously announced.
Why is OpenAI retiring these models?
To streamline the experience, improve performance and reliability, and focus resources on newer models.
Will there be changes to the API?
No changes at this time for the API; continue to monitor official deprecation notices for snapshots and Azure timelines.
How will this affect users?
Most users won’t need to act. The default model selection will shift to current versions, with improved performance and stability.
How can I preserve a legacy model’s tone?
Use Custom Instructions and a short style block (examples of tone, phrasing, and do/don’t rules) to approximate previous behaviour.
Recevez chaque semaine des nouvelles et des conseils sur l'IA directement dans votre boîte de réception
En vous abonnant, vous consentez à ce que Génération Numérique stocke et traite vos informations conformément à notre politique de confidentialité. Vous pouvez lire la politique complète sur gend.co/privacy.
Ateliers et webinaires à venir

Clarté opérationnelle à grande échelle - Asana
Webinaire Virtuel
Mercredi 25 février 2026
En ligne

Travailler avec des coéquipiers IA - Asana
Atelier en personne
Jeudi 26 février 2026
London, UK

From Idea to Prototype - AI in Miro
Virtual Webinar
Weds 18th February 2026
Online
Génération
Numérique

Bureau au Royaume-Uni
33 rue Queen,
Londres
EC4R 1AP
Royaume-Uni
Bureau au Canada
1 University Ave,
Toronto,
ON M5J 1T1,
Canada
Bureau NAMER
77 Sands St,
Brooklyn,
NY 11201,
États-Unis
Bureau EMEA
Rue Charlemont, Saint Kevin's, Dublin,
D02 VN88,
Irlande
Bureau du Moyen-Orient
6994 Alsharq 3890,
An Narjis,
Riyad 13343,
Arabie Saoudite
Numéro d'entreprise : 256 9431 77 | Droits d'auteur 2026 | Conditions générales | Politique de confidentialité
Génération
Numérique

Bureau au Royaume-Uni
33 rue Queen,
Londres
EC4R 1AP
Royaume-Uni
Bureau au Canada
1 University Ave,
Toronto,
ON M5J 1T1,
Canada
Bureau NAMER
77 Sands St,
Brooklyn,
NY 11201,
États-Unis
Bureau EMEA
Rue Charlemont, Saint Kevin's, Dublin,
D02 VN88,
Irlande
Bureau du Moyen-Orient
6994 Alsharq 3890,
An Narjis,
Riyad 13343,
Arabie Saoudite
Numéro d'entreprise : 256 9431 77
Conditions générales
Politique de confidentialité
Droit d'auteur 2026








