AIToday
Large Language ModelsSimon Willison's WeblogPublished: Apr 25, 2026, 14:21 JST

OpenAI releases GPT-5.5 API with new prompting guide—developers must rewrite existing prompts from scratch rather than adapt old ones

3 Key Points

  1. OpenAI made GPT-5.5 available in its API on April 25, 2026, and published a prompting guide warning developers that the new model is not a drop-in replacement for GPT-5.2 or GPT-5.4. The guide explicitly states: 'treat it as a new model family to tune for' and advises starting with fresh prompts instead of carrying over instructions from older versions.

  2. One concrete technique OpenAI highlights: for long-running tasks, send a brief user-visible message acknowledging the request and stating the first step before any tool calls—this prevents users from thinking the AI has crashed. OpenAI's own Codex app already implements this pattern.

  3. For developers maintaining applications built on older GPT models, this means existing prompt optimizations won't automatically work with GPT-5.5—they'll need to re-test and re-tune reasoning effort, verbosity (how much the model explains), tool descriptions, and output format against actual examples. Teams with hundreds of production prompts face weeks of validation work to migrate safely.

  4. To upgrade existing code, developers can run the command `$openai-docs migrate this project to gpt-5.5` in Codex, which provides a migration guide with instructions for rewriting prompts to fit the new model.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Simon Willison's WeblogRead Original Article

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleDeepSeek releases V4 Pro and Flash models with 1M-token context and MIT license, positioning Chinese AI labs as credible open-weight alternatives to closed frontier models