OpenAI is discontinuing six older GPT models starting April 14, 2026, affecting ChatGPT users who rely on legacy code generation tools. The company is retiring GPT-5, GPT-5.1, and GPT-5.2-codex alongside three other Codex variants, while keeping a smaller set of newer and more efficient models available .

Which GPT Models Are Being Retired and Which Survive?

The retirement affects users who have built workflows around older Codex models. OpenAI is keeping four models in active service: GPT-5.4, GPT-5.4-mini, GPT-5.3-codex, and GPT-5.3-codex-spark, which is available exclusively to Pro subscribers . The discontinuation targets models that have been superseded by newer versions with better performance and lower computational costs.

• Models Being Retired: GPT-5, GPT-5.1, GPT-5.2-codex, and three additional Codex variants will no longer be available after April 14
• Models Staying Active: GPT-5.4 and GPT-5.4-mini remain available to all users, while GPT-5.3-codex-spark is reserved for ChatGPT Pro subscribers
• Affected Users: ChatGPT sign-in users who have integrated older Codex models into their development pipelines will need to migrate to supported alternatives

Why Is OpenAI Retiring These Models Now?

Model retirement is a standard practice in AI development as newer versions become available. Maintaining multiple versions of large language models requires significant computational resources and ongoing support. By consolidating to a smaller set of models, OpenAI can focus engineering efforts on improving the remaining options and developing next-generation tools. The company typically retires older models when their performance has been surpassed by newer alternatives that are also more efficient to run .

The timing aligns with broader industry trends toward model consolidation. As AI models become more capable, companies face pressure to streamline their offerings and reduce infrastructure costs. Keeping only the most performant and cost-effective versions allows OpenAI to allocate resources toward research and development of frontier models.

How to Migrate Your ChatGPT Workflows Before April 14

• Audit Your Current Setup: Review which GPT models your applications and scripts currently use by checking your API calls and ChatGPT integration settings
• Test GPT-5.4 Compatibility: Run your existing code against GPT-5.4 and GPT-5.4-mini to ensure they produce acceptable results for your use case before the retirement date
• Update API Calls and Configuration: Modify any hardcoded model references in your codebase to point to the surviving models, testing thoroughly before the April 14 deadline
• Consider Pro Subscription: If your workflow requires GPT-5.3-codex-spark, evaluate whether a ChatGPT Pro subscription is cost-effective for your needs

Users who have not migrated by April 14 may experience errors when attempting to use retired models. OpenAI typically provides advance notice to allow developers time to update their integrations, but the responsibility falls on individual users to make the transition. The company recommends testing new models in a staging environment before deploying changes to production systems.

What This Reveals About OpenAI's Model Strategy

The retirement pattern suggests OpenAI is moving toward a leaner model portfolio focused on performance and efficiency. Rather than maintaining a sprawling catalog of options, the company appears to be consolidating around versions that deliver the best balance of capability and computational cost. This approach mirrors how other major AI labs manage their model releases, prioritizing quality over quantity .

For developers and organizations relying on ChatGPT for code generation, the key takeaway is that model stability cannot be assumed indefinitely. Planning for periodic migrations and staying informed about deprecation timelines helps prevent disruptions to production systems. OpenAI's April 14 deadline provides roughly two weeks for users to complete their transitions, though earlier action reduces the risk of last-minute complications.