OpenAI released GPT-5.5 on Thursday, rolling out the model to ChatGPT Plus, Pro, Business, and Enterprise users along with its Codex coding assistant. The company calls it “our smartest and most intuitive to use model yet,” with benchmark gains concentrated in agentic coding, computer use, and scientific research.

This is a developing story. NCT previously covered GPT-5.5 leaks revealing the model’s codename “Spud” and preliminary capability details. This article covers the official release.

Benchmark Performance

GPT-5.5 scores 82.7% on Terminal-Bench 2.0, which tests complex command-line workflows requiring planning, iteration, and tool coordination, up from GPT-5.4’s 75.1%. On SWE-Bench Pro for real-world GitHub issue resolution, it reaches 58.6%. On OSWorld-Verified for computer use tasks, it hits 78.7%, narrowly edging Claude Opus 4.7’s 78.0%, according to OpenAI.

On FrontierMath Tier 4, the hardest mathematical reasoning tier, GPT-5.5 scores 35.4% compared to Claude Opus 4.7 at 22.9% and Gemini 3.1 Pro at 16.7%.

OpenAI also released a GPT-5.5 Pro variant for Pro, Business, and Enterprise users, which scores higher on BrowseComp (90.1%) and FrontierMath Tier 4 (39.6%).

Efficiency Claims

OpenAI emphasized that GPT-5.5 matches GPT-5.4’s per-token latency while delivering higher performance. The model also uses “significantly fewer tokens” to complete the same Codex tasks, according to the company’s blog post. On Artificial Analysis’s Coding Index, OpenAI claims GPT-5.5 “delivers state-of-the-art intelligence at half the cost of competitive frontier coding models.”

“This model is a real step forward towards the kind of computing that we expect in the future,” OpenAI President Greg Brockman told reporters. “It’s a faster, sharper thinker for fewer tokens compared to something like 5.4.”

Cybersecurity Risk Classification

GPT-5.5 carries OpenAI’s “High” cybersecurity risk classification, meaning it could “amplify existing pathways to severe harm,” though it does not cross the “Critical” threshold, according to CNBC. OpenAI VP of Research Mia Glaese said the company tested the model with internal and external red teamers and nearly 200 trusted early-access partners before release.

The classification arrives weeks after Anthropic limited the rollout of its Mythos model over similar cybersecurity concerns, as reported by TechCrunch.

No API Access Yet

GPT-5.5 is available in ChatGPT and Codex but not yet through OpenAI’s API. The company said API deployments “require different safeguards” and that access would come “very soon.” For enterprises building agentic workflows through the API, the wait continues.

The Release Cadence Question

GPT-5.5 arrives less than two months after GPT-5.4 (released March 5), which itself followed a December release. OpenAI Chief Scientist Jakub Pachocki told TechCrunch the pace should be expected to continue: “We see pretty significant improvements in the short term, extremely significant improvements in the medium term. In fact, I would say, like, I think the last two years have been surprisingly slow.”

Brockman also framed GPT-5.5 as a step toward OpenAI’s planned “super app” combining ChatGPT, Codex, and an AI browser into one unified service. That product has not shipped yet, but the agentic capabilities in GPT-5.5, particularly computer use and multi-tool task completion, suggest the infrastructure is being built model by model.