According to OpenAI, GPT-5.5 stands out as its most proficient agentic coding model yet. It swiftly grasps user intent and executes multi-step tasks with minimal input. This model is designed to carry out diverse functions, including writing and debugging code, conducting research, analyzing data, generating documents and spreadsheets, and even navigating software tools to accomplish tasks from start to finish.
In a media briefing on Thursday (April 23), OpenAI President Greg Brockman highlighted the model’s key strength in its capacity to resolve ambiguous challenges autonomously. He mentioned that it can “examine an unclear issue and determine what actions to take next,” indicating a transformation in how individuals may interact with computers in the future.
Enhanced coding and automation capabilities
A significant enhancement lies in coding performance. GPT-5.5 adeptly manages intricate programming workflows, addresses bugs, tests code, and oversees extensive codebases with improved accuracy. Company benchmarks show significant advancements over its predecessor, GPT-5.4, particularly in resolving real-world coding challenges in a single attempt.
Initial testers, including engineers, have reported enhanced reasoning capabilities, with the model anticipating problems and identifying testing or review needs without explicit instructions.
The model is also optimized for efficiency, yielding higher-quality outputs while consuming fewer tokens for tasks. Despite these enhancements, OpenAI asserts that GPT-5.5 preserves response speeds comparable to earlier versions, effectively managing the typical balance between performance and latency.
Productivity and “computer use” features
Beyond coding, GPT-5.5 is positioned as a resource for daily knowledge work. It can aid in research, produce structured reports, create spreadsheets, and distill complex information. The model is designed to engage more seamlessly with software, assisting users in navigating interfaces, managing workflows, and transitioning between tools more fluidly.
OpenAI has indicated that its Codex system is currently utilized internally across various departments including engineering, finance, and communications to automate workflows. Reports suggest that over 85% of employees are using Codex weekly in areas such as marketing, data science, and product management.
Safety measures
The company noted that GPT-5.5 has been released with improved safety mechanisms. These encompass preparedness evaluations, domain-specific testing, targeted assessments for advanced biology and cybersecurity capabilities, and evaluations carried out with external experts.
Availability and pricing
GPT-5.5 is being made available to Plus, Pro, Business, and Enterprise users across ChatGPT and Codex. The Pro version is currently accessible for Business and Enterprise tiers, with broader API access anticipated soon.
For developers, the standard model is priced at $5 per million input tokens and $30 per million output tokens. The higher-accuracy Pro version is available for $30 and $180, respectively. OpenAI is also providing Batch and Flex pricing options at a 50% discount, along with premium Priority processing.
Although priced higher than GPT-5.4, the company claims that GPT-5.5 is designed to be more efficient overall, delivering superior results while utilizing fewer tokens.
