OpenAI Announces GPT-5.5: What You Need to Know
OpenAI Announces GPT-5.5, Its Most Capable Model Yet
OpenAI on Thursday announced GPT-5.5, its latest artificial intelligence model, describing it as significantly better at coding, operating computers autonomously, and conducting deeper research than its predecessor. The release, which comes just six weeks after GPT-5.4 launched on March 5, 2026, is rolling out immediately to paid ChatGPT subscribers — including Plus, Pro, Business, and Enterprise users — as well as through OpenAI's coding assistant Codex. API access is expected to follow shortly, pending additional cybersecurity safeguards.
Internally codenamed "Spud," GPT-5.5 is described by OpenAI as its first fully retrained large language model since GPT-4.5, and is intended to serve as a technical foundation for a planned "Super App" that would combine ChatGPT, Codex, and a dedicated browser. The announcement came one week after competitor Anthropic launched its own latest model, underscoring what Fortune described as an "extremely fast turnaround that underscores how fiercely frontier AI labs are competing for enterprise customers."
What GPT-5.5 Can Do: Coding, Agentic Tasks, and Office Productivity
According to OpenAI, GPT-5.5 excels across a broad range of productivity tasks: analyzing data, writing and debugging code, operating software, researching online, and creating documents and spreadsheets. Bloomberg reported the model is specifically intended to be better at completing work without much direction, including using email, spreadsheets, calendars, and other applications to follow a user's commands on a computer — a capability often described in the industry as "agentic" AI.
Benchmark figures released alongside the model offer a clearer picture of the performance gains. GPT-5.5 scored 82.7% on Terminal-Bench 2.0, up from 75.1% for GPT-5.4. On OpenAI's internal Expert-SWE coding evaluation, GPT-5.5 achieved 73.1%, compared to 68.5% for its predecessor. On the GDPVal benchmark — which measures performance against human workers — GPT-5.5 outperformed or tied human workers on approximately 85% of tasks, ahead of Anthropic's Opus 4.7 at 80% and GPT-5.4 at 83%, according to Inc. magazine.
OpenAI also tested the model with approximately 200 early-access partners before public release, according to StreetInsider.
GPT-5.5 Pro, Pricing, and Access Tiers
Alongside the base model, OpenAI is releasing GPT-5.5 Pro, a larger version of the model, for users subscribed to the $100-per-month Pro tier or those with a ChatGPT Business or Enterprise account, according to Inc. magazine.
API pricing has been set at $5 per million input tokens and $30 per million output tokens for the standard model. GPT-5.5 Pro carries a significantly higher price point: $30 per million input tokens and $180 per million output tokens, reflecting the compute demands of the larger model variant, according to StreetInsider and Inc.
For enterprises evaluating infrastructure costs, Nvidia — whose GPUs were used to train GPT-5.5, consistent with past OpenAI models — says its new chips can slash the cost of running advanced AI models like GPT-5.5 by up to 35x per token, according to Axios. Nvidia vice president of enterprise computing Justin Boitano said the model can act as a "chief of staff," helping power agents that are already acting as employees at Nvidia.
Safety and Cybersecurity Classifications
OpenAI has classified GPT-5.5's cybersecurity capabilities as "High" under its Preparedness Framework — a meaningful designation, though the company noted the model did not reach the "Critical" threshold, according to StreetInsider. The classification signals that OpenAI views the model as capable enough in cyber-relevant domains to warrant heightened scrutiny, which is part of why API access is being held back temporarily pending "different safeguards," as CNBC reported.
Mia Glaese, OpenAI's vice president of research, addressed the safety process directly in comments reported by CNBC.
"GPT-5.5 underwent extensive third-party safeguard testing and red teaming for cyber and bio [risks], and we've been iterating on our cyber safeguards for months with increasingly cyber capable models." — Mia Glaese, OpenAI Vice President of Research
The Competitive Context: Six Weeks, One Predecessor, One Rival Launch
The pace of GPT-5.5's arrival is striking even by recent standards. GPT-5.4 was released on March 5, 2026 — meaning OpenAI has shipped a fully retrained large language model in roughly six weeks. Fortune characterized that timeline as reflective of how intensely frontier AI labs are competing, particularly for enterprise customers who are increasingly choosing AI platforms as core infrastructure rather than supplementary tools.
The timing also places GPT-5.5 directly in competitive tension with Anthropic, whose latest model launched just one week earlier, according to Axios. OpenAI's decision to release benchmark comparisons against Anthropic's Opus 4.7 — including the GDPVal human-task comparison — suggests the company is actively framing GPT-5.5 as a competitive response in the enterprise segment.
OpenAI also disclosed current scale figures alongside the launch: ChatGPT has more than 900 million weekly active users and over 50 million subscribers, with 9 million paying business users and 4 million active Codex users, according to Fortune. These figures appear positioned to reinforce OpenAI's market standing at a moment of heightened competitive scrutiny.
What OpenAI's President Says About the New Model
Greg Brockman, OpenAI's president and co-founder, has been the primary spokesperson for GPT-5.5's capabilities, emphasizing the model's ability to operate with minimal instruction — a characteristic that distinguishes it from earlier generations of large language models that required detailed prompting to perform well.
"What is really special about this model is how much more it can do with less guidance. It can look at an unclear problem and figure out just what needs to happen next." — Greg Brockman, OpenAI President
"This is a new class of intelligence. It's a big step towards more agentic and intuitive computing." — Greg Brockman, OpenAI co-founder and President
Brockman also described the model in Inc. magazine as "way more intuitive to use" and capable of accomplishing "more with less guidance" — language consistent with OpenAI's broader positioning of GPT-5.5 as a model suited for autonomous, multi-step work rather than single-turn question-and-answer tasks.
What Comes Next for GPT-5.5 and OpenAI
API access for GPT-5.5 is expected to arrive "very soon," according to CNBC, though OpenAI has indicated it requires "different safeguards" before developers can access the model programmatically — a likely reference to the "High" cybersecurity classification under the Preparedness Framework. Developers and enterprises building on the API should expect a brief delay relative to the consumer rollout.
Beyond the immediate release, GPT-5.5 is described by Trending Topics as intended to serve as a technical foundation for OpenAI's planned "Super App" — an integrated product combining ChatGPT, Codex, and a dedicated browser. No timeline for that product has been confirmed in the available reporting.
The model's emphasis on agentic, low-guidance task completion also points toward a broader shift in how OpenAI is positioning its products: less as chatbots that answer questions, and more as autonomous systems that can execute multi-step workflows across real software environments — email, spreadsheets, calendars, code editors, and the open web.
For now, GPT-5.5 is available to paid ChatGPT subscribers and Codex users starting today, April 23, 2026. Businesses and developers looking to evaluate the model at scale will need to wait for API access, which OpenAI says is coming but has not yet dated precisely.
For more tech news, visit our news section.
Why This Matters for Your Productivity
GPT-5.5's core design philosophy — doing more with less guidance, autonomously navigating software, and completing complex multi-step tasks — has direct implications for how individuals and teams manage their time and cognitive load. Whether you're tracking health data in spreadsheets, researching wellness protocols, or managing a packed calendar, AI systems capable of agentic, low-friction assistance are increasingly relevant tools for personal optimization. At Moccet, we're watching this space closely because the intersection of AI capability and human productivity is exactly where meaningful change happens. Join the Moccet waitlist to stay ahead of the curve.