GPT-5.5 by OpenAI

Introducing GPT-5.5: A New Class of Intelligence for Real Work

On April 23, 2026, OpenAI officially announced the release of GPT-5.5, its smartest and most intuitive model yet. This model represents a significant leap forward in the evolution of artificial intelligence, moving beyond simple chat interactions to a new way of getting work done on a computer. GPT-5.5 is designed to understand intent faster and carry out multi-part tasks autonomously, effectively acting as an agent that can plan, use tools, and navigate through ambiguity to finish complex projects.

What’s GPT-5.5?

GPT-5.5 is OpenAI’s latest frontier model, engineered specifically for high-level reasoning and agentic workflows. Unlike its predecessors, GPT-5.5 is not just about generating text; it is built to operate software, move across different tools, and manage the entire lifecycle of a task. It excels at tasks that require reasoning across context and taking action over time, such as debugging code, conducting online research, and creating complex spreadsheets.

Along with the standard model, OpenAI has also introduced GPT-5.5 Pro, a version tailored for even more demanding tasks that require higher accuracy and comprehensive structured responses. GPT-5.5 maintains the same per-token latency as GPT-5.4 while delivering a significantly higher level of intelligence and using fewer tokens to complete tasks, making it both more capable and more efficient.

Key Features of GPT-5.5

1. Advanced Agentic Intelligence

GPT-5.5 can be given a messy, multi-part task and trusted to plan its execution. It navigates through ambiguity, checks its own work, and iterates until the goal is achieved. This makes GPT-5.5 particularly powerful for agentic coding and knowledge work.

2. Industry-Leading Coding Capabilities

In the realm of software engineering, GPT-5.5 has set new benchmarks. On Terminal-Bench 2.0, it achieved a state-of-the-art accuracy of 82.7%. It outperforms previous models in real-world GitHub issue resolution (SWE-Bench Pro) and long-horizon coding tasks. In Codex, GPT-5.5 helps with implementation, refactoring, and system-wide debugging.

3. High Efficiency and Low Latency

Despite being more intelligent, GPT-5.5 matches the real-world serving speed of GPT-5.4. It is more token-efficient, often reaching high-quality outputs with fewer retries and lower token consumption, which reduces costs for developers and enterprise users.

4. Specialized Scientific Research Skills

GPT-5.5 shows remarkable gains in scientific and technical research. It can explore ideas, gather evidence, and interpret results in fields like genetics and quantitative biology. It even contributed to discovering a new proof about Ramsey numbers in the field of combinatorics.

5. Enhanced Computer Use and Tool Navigation

Combined with Codex, GPT-5.5 can see what is on a screen, click, type, and navigate interfaces with precision. This allows the model to work alongside humans in real-time professional environments.

Use Cases for GPT-5.5

Professional Coding and Development

For senior engineers, GPT-5.5 serves as a high-intelligence partner. It can re-architect systems, catch potential bugs before they occur, and handle complex merge conflicts. Users at companies like NVIDIA and Cursor report that GPT-5.5 stays on task significantly longer and shows "conceptual clarity" that previous models lacked.

Knowledge Work and Business Automation

In business settings, GPT-5.5 automates document generation, spreadsheet modeling, and data synthesis. For example, OpenAI’s finance team used GPT-5.5 in Codex to review over 24,000 tax forms, accelerating the task by two weeks. Marketing and communication teams use it to analyze data and build automated agents for request handling.

Scientific and Academic Research

Researchers use GPT-5.5 Pro as a research partner to critique manuscripts, stress-test technical arguments, and analyze massive datasets. In immunology, it has been used to analyze gene-expression datasets with tens of thousands of genes, producing reports that would normally take months of human effort.

Cybersecurity and Defense

GPT-5.5 is an incremental step toward solving tough cybersecurity challenges. It is being used to find and patch security vulnerabilities. Through the Trusted Access for Cyber program, verified defenders can use GPT-5.5 to secure critical infrastructure and protect public data systems.

Performance Benchmarks

GPT-5.5 and GPT-5.5 Pro have been tested against a wide variety of industry benchmarks, consistently outperforming competitors like Claude Opus 4.7 and Gemini 3.1 Pro.

• Terminal-Bench 2.0: 82.7% (GPT-5.5) vs 69.4% (Claude Opus 4.7)
• GDPval (Knowledge Work): 84.9% (GPT-5.5)
• OSWorld-Verified (Computer Use): 78.7% (GPT-5.5)
• BixBench (Bioinformatics): 80.5% (GPT-5.5)
• Tau2-bench Telecom: 98.0% (GPT-5.5)
• ARC-AGI-2 (Abstract Reasoning): 85.0% (GPT-5.5)

Safety and Safeguards

OpenAI has deployed GPT-5.5 with its strongest set of safeguards to date. The model was evaluated across safety and preparedness frameworks, including targeted testing for cybersecurity and biology capabilities.

• Cybersecurity Protections: Stricter classifiers have been implemented to limit potential abuse while preserving access for beneficial defensive work.
• Preparedness Framework: Biological, chemical, and cybersecurity capabilities are treated as "High" risk categories, requiring robust monitoring and authenticated usage.
• Redteaming: Feedback was collected from nearly 200 trusted early-access partners to refine the model's safety profile before the broad release.

How to Use GPT-5.5

GPT-5.5 is available through multiple platforms depending on your user tier:

1. ChatGPT: Plus, Pro, Business, and Enterprise users can access GPT-5.5 Thinking for smarter, more concise answers to complex professional problems.
2. GPT-5.5 Pro in ChatGPT: Available to Pro, Business, and Enterprise users for even more difficult tasks requiring comprehensive and highly accurate responses.
3. Codex: Available for Plus, Pro, Business, Enterprise, Edu, and Go plans with a 400K context window. It includes a "Fast mode" for generating tokens 1.5x faster.
4. API: Developers will soon have access to gpt-5.5 and gpt-5.5-pro via the Responses and Chat Completions APIs with a 1M context window.

FAQ

What is the difference between GPT-5.5 and GPT-5.5 Pro?

GPT-5.5 is the standard high-intelligence model for general work, while GPT-5.5 Pro is designed for the hardest questions and highest-accuracy tasks, offering more comprehensive and well-structured responses in specialized fields like law, education, and data science.

How much does the GPT-5.5 API cost?

For developers, gpt-5.5 is priced at $5 per 1M input tokens and $30 per 1M output tokens. The gpt-5.5-pro model is priced higher at $30 per 1M input tokens and $180 per 1M output tokens to reflect its higher accuracy.

Is GPT-5.5 faster than GPT-5.4?

While GPT-5.5 provides a higher level of intelligence, it matches the per-token latency of GPT-5.4 in real-world serving. Furthermore, it is more token-efficient, meaning it can often complete the same task using fewer tokens.

Can GPT-5.5 be used for cybersecurity?

Yes, but with strict safeguards. OpenAI offers Trusted Access for Cyber for verified users to use GPT-5.5 for defensive purposes, such as securing code and protecting critical infrastructure, while implementing classifiers to prevent malicious use.