According to a company statement, GPT-5.5 excels at writing and debugging code, online research, data analysis, creating documents and spreadsheets, operating software, and switching between tools to complete a task.
Thus, the user can assign GPT-5.5 a complex, multi-part task, and the model “will plan, use tools, review its work, navigate ambiguous situations, and move forward.”
The company emphasizes that GPT-5.5 understands the task the user wants to perform faster and can carry out much of the work itself, making it more efficient in problem solving.
“Improvements are especially notable in agent coding, computer use, and early scientific research, areas where progress depends on contextual reasoning and decision making over time,” he notes.
Additionally, GPT-5.5 equalizes latency per tokenthat is, the response speed per unit of text, of one of its predecessors, GPT-5.4.
The new model is rolling out to Plus, Pro, Business and Enterprise users of ChatGPT and Codex, a cloud-based enterprise software engineering agent.
However, the company works with partners and customers to meet necessary security requirements and will not currently release this technology as an API (application programming interface), which would allow other companies to integrate it into their applications.
OpenAI’s announcement comes weeks after Anthropic, one of its competitors, announced Claude Mythos Preview, a new computing model with programming and cybersecurity capabilities so advanced that the company decided to veto its launch to the general public, considering it a risk to global security.
Companies that can access this model include Amazon, Apple, Google, Microsoft, Nvidia, CrowdStrike, JPMorgan Chase, Cisco, Broadcom, Palo Alto Networks and the Linux Foundation.
Like Anthropic, OpenAI also decided to share one of its previous models, GPT-5.4-Cyber, with a select group of partners.