OpenAI has launched GPT 5.5, describing it as its smartest and most intuitive model yet, with major gains in coding, research, data analysis, document creation, and computer based task execution.

The company said the model is rolling out starting today to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, while API access is expected soon.

Introducing GPT-5.5

A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done.

Now available in ChatGPT and Codex. pic.twitter.com/rPLTk99ZH5

— OpenAI (@OpenAI) April 23, 2026

The release positions GPT 5.5 as a step up from GPT 5.4 in areas tied to agentic work. OpenAI said the model is better at understanding user intent, planning multi step tasks, using tools, checking its own work, and navigating ambiguity with less back and forth.

The company also said GPT 5.5 matches GPT 5.4 on per token latency in real world serving while using fewer tokens on many coding tasks.

OpenAI highlighted particularly strong gains in coding benchmarks. GPT 5.5 scored 82.7% on Terminal Bench 2.0, up from 75.1% for GPT 5.4, and 73.1% on Expert SWE, compared with 68.5% for the prior model. On SWE Bench Pro, GPT 5.5 reached 58.6%, narrowly ahead of GPT 5.4’s 57.7%.

The company also framed GPT 5.5 as a stronger model for knowledge work and computer use. OpenAI said it scored 84.9% on GDPval, 78.7% on OSWorld Verified, and 98.0% on Tau2 bench Telecom without prompt tuning, reflecting gains in tasks such as information synthesis, spreadsheet modeling, workflow execution, and operating software environments.

In scientific and technical work, OpenAI said GPT 5.5 improved over GPT 5.4 on benchmarks including GeneBench and BixBench, and pointed to early use cases in genetics, mathematics, and biomedical research.

The company also said an internal version of the model helped discover a new proof related to Ramsey numbers that was later verified in Lean.

OpenAI said GPT 5.5 was released with what it described as its strongest safeguards to date. The company said it conducted preparedness evaluations, external red teaming, and targeted testing for advanced cybersecurity and biology capabilities, and classified the model’s cyber and biological or chemical capabilities as High under its Preparedness Framework. OpenAI also launched a GPT 5.5 Bio Bug Bounty alongside the release.

For developers, OpenAI said GPT 5.5 will soon be available through the Responses API and Chat Completions API at $5 per 1 million input tokens and $30 per 1 million output tokens, with a 1 million context window. The company said gpt 5.5 pro will also be added to the API at $30 per 1 million input tokens and $180 per 1 million output tokens.

The release underscores OpenAI’s push to make ChatGPT and Codex more capable at carrying out real computer work from end to end. Rather than focusing only on chatbot responses, the company is pitching GPT 5.5 as a model built to execute across tools, software, and long running workflows with more autonomy and fewer retries.