OpenAI retires GPT-4o from ChatGPT

OpenAI announced on 2026-01-29 that on 2026-02-13 it would retire GPT-4o, GPT-4.1, GPT-4.1 mini and o4-mini from ChatGPT (with GPT-5 Instant and Thinking, retirement previously announced); the API…

Date
13 February 2026
Who
OpenAI
Confidence
High (dates and OpenAI's stated reasons); Medium (the sycophancy ranking, which
Deep dive
RLHF and instruction tuning (how base models became assistants)

Tier: Supporting · Significance: 3/5 · Org(s): OpenAI · Confidence: High (dates and OpenAI's stated reasons); Medium (the sycophancy ranking, which is press-reported) OpenAI announced on 2026-01-29 that on 2026-02-13 it would retire GPT-4o, GPT-4.1, GPT-4.1 mini and o4-mini from ChatGPT (with GPT-5 Instant and Thinking, retirement previously announced); the API was unchanged at the time. OpenAI said GPT-4o deserved special context. It had first deprecated the model, then restored access during the GPT-5 release after Plus and Pro users asked for more time and for its conversational warmth, and the feedback shaped GPT-5.1 and 5.2. The stated reason for retiring it was that most usage had moved to GPT-5.2, with only 0.1% of users still choosing GPT-4o each day (OpenAI, 2026-01-29, archived copy; OpenAI help center, archived copy, which adds that Business, Enterprise and Edu customers kept GPT-4o in Custom GPTs until 2026-04-03). TechCrunch reported that GPT-4o ranked highest for sycophancy among OpenAI's models and that, at 0.1%, the remaining users still number roughly 800,000, with thousands of users protesting; OpenAI's own post does not use the word sycophancy (I searched the archived text) (TechCrunch, 2026-02-13). It matters here because it ends the sequence that began with the April 2025 rollback (B05-41). A model whose flattery had been traced to approval-based rewards was restored after users objected to its first retirement, then withdrawn, with the lawsuits as backdrop (B05-42b). The TechCrunch piece mentions multiple suits without a count, and I did not verify the "13 consolidated suits" figure that some summaries give. Sources: OpenAI blog · OpenAI help center · TechCrunch

Read it in the deep dive