OpenAI has introduced GPT-5.5, positioning it as its most capable and intuitive model yet, with a focus on helping users complete complex, multi-step tasks more independently.

The release marks a continued push toward “agentic” AI systems that can plan, execute, and refine work with minimal human intervention.

The company said the model improves how users interact with AI across coding, research, and general knowledge work.

Instead of guiding every step, users can now assign broader tasks and rely on the model to navigate ambiguity and complete workflows.

“GPT-5.5 understands what you’re trying to do faster and can carry more of the work itself,” the company stated.

Stronger agentic coding

GPT-5.5 shows major gains in coding, especially in complex workflows that require planning and tool coordination.

On Terminal-Bench 2.0, it achieved 82.7% accuracy, a state-of-the-art score.

On SWE-Bench Pro, it reached 58.6%, solving more real-world GitHub issues in a single pass than earlier versions.

The model also outperformed its predecessor in long-horizon engineering tasks measured by internal benchmarks.

These tasks often take human developers up to 20 hours to complete.

OpenAI said the improvements go beyond benchmarks. Early testers reported that GPT-5.5 better understands system architecture and failure points.

It can identify where fixes belong and predict downstream impacts across a codebase.

The company emphasized efficiency alongside capability. GPT-5.5 matches GPT-5.4’s per-token latency despite higher intelligence.

It also uses fewer tokens to complete the same tasks, lowering computational cost.

“GPT-5.5 delivers this step up in intelligence without compromising on speed,” OpenAI noted. It added that the model performs at a higher level while maintaining real-world responsiveness.

Expanding real-world use

Beyond coding, GPT-5.5 expands its role in everyday knowledge work.

The model can move across tasks such as gathering information, analyzing data, and generating structured outputs like documents and spreadsheets.

OpenAI said this reflects a broader shift toward AI systems that can actively operate software and tools.

The model can interpret interfaces, take actions, and transition between workflows with minimal friction.

Internal adoption highlights these capabilities.

More than 85% of OpenAI employees now use Codex weekly across departments, including engineering, finance, and marketing.

In one example, the communications team used GPT-5.5 to process six months of speaking request data.

The system built a scoring and risk framework and helped automate low-risk approvals.

In finance, the model reviewed 24,771 K-1 tax forms totaling over 71,000 pages.

The workflow excluded personal data and reduced processing time by two weeks.

Another team automated weekly business reporting, saving between five and ten hours each week.

OpenAI also stressed safety in the rollout.

The company said it deployed its strongest safeguards so far, including red-teaming, advanced testing, and feedback from nearly 200 early-access partners.

“Today, GPT-5.5 is rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex,” the company said.

API access will follow after additional safety and scaling requirements are met.

The launch signals OpenAI’s continued focus on building infrastructure for agentic AI.

The company aims to expand how people and businesses use AI to complete complex work across domains.