GPT-5.5: Smarter decisions, Less guidance required
GPT- 5.5 is OpenAI's most capable model designed for complex, real-world work. It figures out what needs to happen next, uses tools to get there, and checks its own work along the way.
Trusted by users from 10,000+ companies
Frontier Performance in Real-world Tasks
GPT-5.5 leads across the evaluations that mirror how professionals and researchers actually work.
Excels at Clinician-use Evaluation
GPT-5.5 scores higher than any previous GPT model on HealthBench Professional,designed for clinical use cases, with both higher raw and length-adjusted scores.
Top Agentic Coding Model
GPT-5.5 leads SWE-Bench Verified, placing it at the top of the field for resolving real production software issues autonomously across full codebases.
Most Accurate GPT Model
On conversations where users had previously flagged factual errors, GPT-5.5's individual claims are 23% more likely to be correct, improving precision and trustworthiness.
Work that Figures Itself Out
GPT-5.5 understands what a task actually requires before starting, identifies the gaps in ambiguous briefs, and builds a plan without waiting to be walked through it. Early access teams reported saving up to 10 hours of work per week.
Additional Capabilities GPT-5.5 Brings
Apart from headline changes, GPT-5.5 offers smaller improvements that compound into bigger advantages.
Token efficiency
GPT-5.5 completes the same tasks using significantly fewer tokens than GPT-5.4, delivering more capable results at lower cost in Codex.
Same-speed serving
Despite the jump in capability, GPT-5.5 matches GPT-5.4's per-token latency in real-world serving. More intelligence, no wait penalty.
Computer use
GPT-5.5 can operate software directly, navigating interfaces, filling forms, and executing multi-step tasks the way a person would.
Scientific reasoning
Meaningful gains on scientific and technical research workflows, including early-stage drug discovery and experimental debugging tasks.
Destructive action prevention
GPT-5.5 distinguishes its own work from user work and reverts only its own changes, protecting user output even in complex, multi-agent environments.
Context retention
Holds context reliably across large systems and long sessions, carrying changes through surrounding codebases without losing track of earlier decisions.
Document creation
Improved latency on document and spreadsheet creation tasks, producing complete, structured outputs faster without sacrificing accuracy.
Cybersecurity access
Verified defenders can apply for Trusted Access for Cyber to unlock GPT-5.5's advanced security capabilities with fewer restrictions for legitimate defensive work.
Codex integration
GPT-5.5 is the default model in Codex for all paid tiers, with OpenAI's tuning specifically optimised for agentic coding performance at lower token cost.
Frequently Asked Questions
Quick answers covering access, pricing, capabilities, and how GPT-5.5 compares to previous models.