#OpenAIReleasesGPT-5.5 OpenAI Launches GPT-5.5: A New Era of Agentic AI Ushers in the ‘Super App’ Ambition

San Francisco – In a surprise move that has reset the competitive landscape of artificial intelligence, OpenAI has officially launched GPT-5.5.

Released just six weeks after GPT-5.4, the new model—codenamed "Spud"—is not merely an incremental upgrade. OpenAI President Greg Brockman described it as “a new class of intelligence for real work,” signaling a fundamental shift from conversational chatbots to autonomous "agentic" systems capable of performing complex, multi-step tasks without hand-holding .

A Focus on “Doing,” Not Just “Chatting”

Unlike its predecessors, GPT-5.5 is architected as an Agentic Runtime. It is designed to plan, execute, verify, and iterate across digital tools—such as code interpreters, browsers, and office software—to finish a job with minimal supervision .

Early testers have noted a significant leap in "conceptual clarity" and persistence. In one notable case, the model successfully merged a complex branch with hundreds of frontend changes into a modified main branch in a single 20-minute attempt—a task previous models failed to complete .

Benchmarks and Pricing

The performance gains are immediate. On Terminal-Bench 2.0, which tests complex command-line workflows, GPT-5.5 scored 82.7%, besting Claude Opus 4.7’s 69.4% . It also leads in knowledge work with an 84.9% score on GDPval.

However, the power comes at a price. OpenAI has doubled the cost: Standard GPT-5.5 is priced at **$5 per million input tokens** and $30 per million output tokens, while GPT-5.5 Pro (for heavy-duty tasks) costs $30/$180. OpenAI justifies this by noting the model uses "significantly fewer tokens" to complete tasks, lowering the total cost of ownership .

The ‘Super App’ Strategy and Safety

The launch is a cornerstone of CEO Sam Altman’s strategy to build an AI "Super App," integrating ChatGPT, Codex, and an AI browser into a unified platform for enterprise work .

Given its capabilities, OpenAI has placed GPT-5.5 in the "High" risk category for cybersecurity, noting it could amplify existing misuse pathways. The company underwent extensive red-teaming and external safety testing before rollout .