OpenAI released GPT-5.5 on Thursday, its most advanced model yet, arriving just seven days after Anthropic unveiled its latest offering in an accelerating race to capture business customers.
The model, codenamed "Spud," processes information faster while consuming fewer computational tokens than its predecessor GPT-5.4, according to the company. More critically, it can tackle multi-step workflows with minimal user direction, handling complex tasks autonomously without requiring detailed step-by-step instructions.
"This is a new class of intelligence," OpenAI co-founder Greg Brockman said during a briefing with reporters. "It's a big step towards more agentic and intuitive computing."
OpenAI engineered GPT-5.5 to excel in areas where sustained reasoning and task execution matter most: coding, computer use, office productivity and early-stage scientific research. Early testers reported saving up to 10 hours weekly on work tasks, with the model functioning as a kind of autonomous assistant that can plan, deploy tools and verify results without constant human intervention.
The rollout targets paid ChatGPT and Codex users starting Thursday, with API access coming once OpenAI integrates additional security measures. Performance speed in real-world scenarios matches GPT-5.4, the company says.
Nvidia's computing infrastructure powers GPT-5.5, and the chip maker is making it a centerpiece of its enterprise pitch. Justin Boitano, Nvidia's vice president of enterprise computing, told Axios the model can serve as a "chief of staff" for agents already functioning as workers at Nvidia. The company's newest chips reduce per-token costs for running advanced AI systems like GPT-5.5 by as much as 35 times, a significant advantage for enterprises managing AI expenses.
The timing reflects shifting competitive dynamics. OpenAI executives previously characterized Anthropic's advances as a "code red" moment, spurring a strategic pivot toward enterprise adoption. Nvidia and OpenAI have coordinated to create what Boitano described as "a blueprint to make it easier for every enterprise to roll this out."
The release underscores a broader transformation in AI deployment. As models take on increasingly complex, time-consuming tasks, the economics of running them become as significant as their raw capabilities. Brockman framed this shift as a move toward "a compute-powered economy," where economic value flows to whoever controls computational capacity.
Author James Rodriguez: "OpenAI is betting the farm on enterprise stickiness with this one, and pairing speed gains with cost efficiency is exactly the move needed to keep Anthropic from stealing the business playbook."
Comments