OpenAI ships GPT-5.5—a step toward the AI super-app
What is new
On April 23, 2026 OpenAI announced GPT-5.5, a new generation of its flagship model. Highlights:
- Better coding — higher SWE-bench scores, improved multi-file refactoring
- Computer use — the model controls GUI and browser more reliably and finishes tasks
- Deep research — capability to autonomously run long research tasks
- Workspace agents — admin role-based controls, free until May 6, 2026
Availability
The model is rolling out to Plus, Pro, Business and Enterprise plans in ChatGPT and Codex.
Why it matters
GPT-5.5 is a step toward a so-called “super-app” a single interface combining chat, code and browsing. For dev teams it means less context switching and the ability to delegate larger multi-step tasks.
My take
For me, this is a signal that agentic workflows are graduating from research preview to production. Clients who were waiting for “grown-up” tooling now have the green light.
May 28, 2026 update: Instant is cleaner, canvas is being phased out
OpenAI updated GPT-5.5 Instant in ChatGPT and the API on May 28, 2026. This is not a new model and it does not need a separate article, but it matters in daily use: responses should be more natural, less verbose and less dependent on bullet-heavy formatting.
OpenAI also notes that canvas is no longer available for GPT-5.5 Instant and GPT-5.5 Thinking. Writing and code work is moving directly into responses through writing blocks and code blocks. Paid users can still access canvas temporarily through legacy models until those models are retired.
The practical takeaway is simple: if your internal guides, training or workflows still point users to a specific older ChatGPT model, it is time to update them around GPT-5.5 Instant / Thinking.
Sources: OpenAI ChatGPT release notes and OpenAI model release notes.