OpenAI unveils GPT-5.5 for complex AI-driven tasks, strengthens ‘super app’ roadmap

Sam Altman-led OpenAI has released its new generation AI model, the GPT-5.5, that brings upgrades in writing, debugging code, researching online, analysing data, and more. The company says it's the “smartest and most intuitive to use model” yet, as it comes with increased capabilities and upgrades. 

OpenAI co-founder and president Greg Brockman also stated the model launch brings the company one step closer to building a “super app,” and that it comes with a major advancement in “agentic and intuitive computing.”

Must read: Infosys bring OpenAI’s AI coding and automation tools to its platform: What it means for engineers

OpenAI’s GPT-5.5: What’s new and upgraded

According to OpenAI’s press note, the GPT-5.5 can handle complex real-world tasks with minimal human guidance. The model goes beyond simple question-answering and instead manages long, multi-step workflows across coding, research, and productivity tools. OpenAI says you can use the model to “plan, use tools, check its work, navigate through ambiguity, and keep going.” 

The company claims that the model comes with advanced agentic coding, computer use, knowledge work, and scientific research capabilities, where reasoning is essential. It also claims that the model is more autonomous than its predecessor, making it more capable for real-world use.

Due to GPT-5.5’s agentic capabilities, the company claims that it is particularly useful for enterprise workflows, including engineering, finance, legal analysis, and scientific research.

Must read: Sam Altman's outside bets raise fresh conflict questions as OpenAI nears IPO

GPT-5.5 capabilities and performance

OpenAI claims that the GPT-5.5 model comes with improvements in coding and software engineering-related tasks. On Terminal-Bench 2.0, the model achieved an accuracy of 82.7%. On SWE-Bench Pro, it reaches 58.6%. 

It also comes with improvements in token efficiency, meaning the model can complete tasks using fewer computational resources, making it faster and more cost-efficient. GPT-5.5 also understands, “why something is failing, where the fix needs to land, and what else in the codebase would be affected.”

In terms of knowledge work, GPT-5.5 reaches 78.7% in the OSWorld-Verified benchmark and 98.0% in Tau2-bench Telecom without prompt tuning.

GPT-5.5 availability 

OpenAI's latest AI model, the GPT-5.5, is rolling out to GPT‑5.5 is rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex. GPT‑5.5 Pro will be available to Pro, Business, and Enterprise users in ChatGPT. The company also assured that it's bringing GPT‑5.5 and GPT‑5.5 Pro to the API soon.