OpenAI on Thursday launched its most succesful AI system, GPT-5.5, which the corporate says will allow a extra highly effective Codex coding agent. OpenAI is fast to say, nonetheless, that GPT-5.5 will energy the widening set of basic digital work duties that Codex is able to. The system is considerably higher than earlier releases at serving to with scientific work, together with inventive facets of producing new hypotheses and testing them.
The system represents an enchancment in autonomous or agentic functionality. GPT-5.56 “represents a step towards AI techniques that may full advanced, multi-step duties on a pc with out human steering,” OpenAI says in a weblog publish.
GPT-5.5, OpenAI says, scores higher than every other AI system on the Terminal-Bench 2.0 benchmark, which assessments for advanced command-line workflows requiring planning, iteration, and power coordination. It scores 82.7%, in comparison with 75.1% for GPT-5.4, 69.4% for Anthropic’s Opus 4.7, and 68.5% for Google’s Gemini 3.1 Professional. On OSWorld-Verified, which measures whether or not a mannequin can function a pc independently, it scores 78.7% versus 75.0% for GPT-5.4. Anthropic’s Opus 4.7 scored 78%.
Anthropic’s latest frontier mannequin, referred to as Mythos, is completed however has not been publicly launched.
OpenAI says it’s seen a surge in customers for its Codex coding agent, with about 4 million builders utilizing the instrument weekly. Throughout a name with press, OpenAI CEO Greg Brockman stated GPT-5.5 will allow Codex to provide polished code, and go about coding tasks with the judgement of a senior software program engineer.
OpenAI is positioning GPT-5.5 as its strongest agentic coding mannequin. On SWE-Bench Professional, a benchmark measuring real-world GitHub difficulty decision, it resolves 58.6% of duties end-to-end in a single cross. Builders who examined the system early stated GPT-5.5 has a greater understanding of the “form” of a software program system, and might higher perceive why one thing is failing, the place the repair is required, and what else within the codebase could be affected.
The GPT-5.5 launch comes simply weeks after the discharge of the GPT-5.4 system. OpenAI is below stress to maintain tempo with its rival Anthropic–particularly within the space of AI coding assistants, which have proved to be AI’s greatest impression on enterprise operations up to now. The tempo of system releases is accelerating as a result of the AI itself is doing numerous the heavy lifting of coding the AI techniques themselves.
Some have speculated that the brand new OpenAI mannequin could be as massive as 10 trillion parameters, however Brockman declined to say.
GPT-5.5 is rolling out to OpenAI’s Plus, Professional, Enterprise, and Enterprise subscription customers in ChatGPT and Codex. GPT-5.5 Professional, a higher-accuracy model, is offered to Professional, Enterprise, and Enterprise customers.
