GPT-5.5 Arrives: OpenAI's Sharpest Thinker Yet Takes Aim at Autonomous Computing

News Summary

OpenAI on Thursday, April 23, 2026 (EDT), officially released GPT-5.5, codenamed "Spud," its most capable AI model to date, rolling it out to paid ChatGPT and Codex subscribers. The new model delivers significant advances in agentic coding, autonomous computer use, scientific research workflows, and multi-step task execution — all while matching the response speed of its predecessor GPT-5.4. GPT-5.5 is described by OpenAI co-founder Greg Brockman as "a new class of intelligence" and a meaningful step toward more autonomous, intuitive computing. API access is expected to follow shortly after additional cybersecurity safeguards are finalized.

Model Capabilities and Design Philosophy

GPT-5.5 is designed as a faster, sharper thinker that handles multi-step workflows more autonomously with less user input, representing a notable shift in how AI models approach complex, open-ended tasks. Rather than requiring users to guide the model step by step, GPT-5.5 excels at analyzing data, writing and debugging code, operating software, researching online, and creating documents and spreadsheets. OpenAI President Greg Brockman highlighted that the model's most distinctive quality is its ability to work with minimal direction: "It can look at an unclear problem and figure out just what needs to happen next."

Mark Chen, chief research officer at OpenAI, said that GPT-5.5 was better at navigating computer work than its predecessors and that the model shows meaningful gains on scientific and technical research workflows, including potential contributions to drug discovery.

Availability and Rollout

GPT-5.5 is widely available starting Thursday (EDT), deploying to Plus, Pro, Business, and Enterprise users in ChatGPT, while GPT-5.5 Pro is headed to Pro, Business, and Enterprise users. API access is expected to arrive "very soon," but those deployments require different safeguards. The model was trained on Nvidia GPUs, consistent with prior OpenAI models, and Nvidia employees will gain access after an internal subset tested the model for several weeks.

Benchmark Performance and Platform Context

According to OpenAI, GPT-5.5 consistently scores higher than its previous models, as well as models from other frontier AI labs, across a range of benchmarks. The company also reported 4 million active Codex users, 9 million paying business users on ChatGPT, over 900 million weekly active users, and more than 50 million subscribers.

Enterprise adoption is already underway. The Bank of New York has been testing GPT-5.5 in recent weeks. CIO Leigh-Ann Russell noted: "What we're actually seeing from 5.5, that I think is really important for a highly regulated institution, is the response quality — but also a really impressive hallucination resistance."

Agentic Computing and the Road to a Super App

Brockman described GPT-5.5 as a big step "towards more agentic and intuitive computing," and noted it brings the company one step closer to the creation of OpenAI's "super app." Teams with early access were able to gut-check vibe-coded work, review thousands of additional documents, and save up to 10 hours of work per week. Nvidia's vice president of enterprise computing Justin Boitano noted that the model can act as a "chief of staff," powering agents that are already acting as employees at Nvidia.

Safety and Cybersecurity Considerations

GPT-5.5 underwent extensive third-party safeguard testing and red teaming for cyber and bio risks, with OpenAI iterating on its cyber safeguards for months with increasingly capable models. Mia Glaese, OpenAI's vice president of research, stated that the company has "a strong and longstanding strategy" for its approach to cybersecurity and has refined a durable approach to rolling out models safely.

Development Pace

This release caps a blistering development pace: GPT-5 in August 2025, GPT-5.2 in November, GPT-5.4 in January 2026, and now GPT-5.5 — all within eight months. The release coming just six weeks after GPT-5.4 underscores how frontier AI labs are competing for enterprise customers and how models are increasingly evolving through continuous, incremental updates.