gpt-4-32k vs gpt-4o
OpenAI names gpt-4o as the replacement for gpt-4-32k. gpt-4-32k shut down on Jun 6, 2025. At standard list prices, gpt-4o costs -94% per token (blended 3:1 input:output).
Retired→Active
Per 1M tokensgpt-4-32kgpt-4o
Input$60$2.50
Output$120$10
Blended change-94%
ProviderOpenAIOpenAI
Statusretiredactive
ShutdownJun 6, 2025—
Context window——
Migrating
- Replace the ID
gpt-4-32kwithgpt-4oeverywhere it appears, including .env files, config and infrastructure templates. - Run your evaluation set or golden prompts on both before switching: output style, refusals and tool calling can differ between generations.
- Check max output tokens and parameters: newer models sometimes rename or drop sampling options.
- Watch the repo on ModelRot so the next retirement doesn't surprise you. gpt-4o lifecycle.