OpenAI rolls out GPT-5.5 to ChatGPT, Codex users

On April 24, OpenAI released GPT-5.5 in ChatGPT and Codex for Plus, Pro, Business, and Enterprise users as its latest model for coding, research, and other computer-based tasks.

The company said API access will come later due to additional safety and security safeguards at scale.

OpenAI said GPT-5.5 beat GPT-5.4 on coding and knowledge work tests, including Terminal-Bench 2.0, SWE-Bench Pro, and GDPval, while matching its per-token latency.

The model was reviewed under its safety framework with internal and external red team testing and feedback from nearly 200 early access partners before release.

🔗 Source: OpenAI

🧠 Food for thought

Implications, context, and why it matters.

OpenAI’s benchmark choices center on coding and knowledge work

OpenAI’s GPT-5.5 claims lean on specialized tests built for coding tasks and professional knowledge work.
SWE-Bench Pro and Terminal-Bench 2.0 track how well a model solves software engineering problems and manages complex command-line workflows. GDPval measures professional knowledge work across occupations 1.
That emphasis, rather than broader creative or general-purpose tests, suggests the model is aimed first at paid users on Plus, Pro, Business, and Enterprise.

Delayed API access fits a safety-first rollout

OpenAI is delaying application programming interface (API) access for GPT-5.5 because wider deployment needs more safety and security safeguards at scale. That differs from an API-first launch, though OpenAI does not describe it as a broader strategy change.
Launching first in ChatGPT and Codex lets the company watch real-world use among paid customers before opening the model through the API.
This sequence may give OpenAI’s own products an early edge over its developer ecosystem. Still, the available facts do not support stronger claims about a new industry standard or slower business innovation.