OpenAI Just Dropped GPT-5.5 and It Wants Your Whole Workflow

OpenAI on Thursday released GPT-5.5, its newest AI model, which the company calls its "smartest and most intuitive to use model" yet. The release lands just six weeks after the company debuted GPT-5.4, an extremely fast turnaround that underscores how fiercely frontier AI labs are competing for enterprise customers.

I've been tracking this release cycle closely, and the cadence is starting to feel less like product launches and more like rolling firmware updates. The pitch this time is agentic work: giving the model a messy task and trusting it to finish.

What GPT-5.5 Actually Does

According to OpenAI's announcement, GPT-5.5 excels at writing and debugging code, researching online, analyzing data, creating documents and spreadsheets, operating software, and moving across tools until a task is finished. Instead of carefully managing every step, you can give GPT-5.5 a messy, multi-part task and trust it to plan, use tools, check its work, navigate through ambiguity, and keep going.

The model is internally codenamed "Spud", and per trendingtopics, it's reportedly OpenAI's first fully retrained large language model since GPT-4.5.

Benchmarks and Performance

Per Fast Company, GPT-5.5 scores better than any other AI system on the Terminal-Bench 2.0 benchmark, which tests for complex command-line workflows requiring planning, iteration, and tool coordination. It achieved a rating of 82.7%, showing a clear lead over GPT-5.4's score of 75.1%, Anthropic's Opus 4.7's score of 69.4%, and Google's Gemini 3.1 Pro's score 68.5%.

On real-world coding tasks, on SWE-Bench Pro, a benchmark measuring real-world GitHub issue resolution, it resolves 58.6% of tasks end-to-end in a single pass. Computer use numbers also moved: on OSWorld-Verified, it scores 78.7%, versus 75% for GPT-5.4 and 78% for Anthropic's Opus 4.7.

Notably, OpenAI says GPT-5.5 matches GPT-5.4's response speed in real-world use despite the capability jump.

Pricing and Availability

For API developers, gpt-5.5 will soon be available in the Responses and Chat Completions APIs at $5 per 1M input tokens and $30 per 1M output tokens, with a 1M context window. The Pro tier is steeper: gpt-5.5-pro will be released in the API for even higher accuracy, priced at $30 per 1M input tokens and $180 per 1M output tokens.

In ChatGPT, GPT-5.5 is rolling out to OpenAI's paid subscribers, including its Plus, Pro, Business, and Enterprise users, in ChatGPT and its coding assistant Codex on Thursday.

The Enterprise Play

The subtext here is Anthropic. Fortune reports that the Bank of New York has been testing GPT-5.5 in recent weeks, alongside early access to models from rivals like Anthropic. CIO Leigh-Ann Russell said the improvements are meaningful, pointing to response quality and impressive hallucination resistance.

On the infrastructure side, Codex is now powered by GPT-5.5, which runs on NVIDIA GB200 NVL72 rack-scale systems, with GB200 NVL72 capable of delivering 35x lower cost per million tokens compared with prior-generation systems.

Final Thoughts

The Terminal-Bench jump from 75.1% to 82.7% is the number I keep circling back to. Command-line agentic work is where coding assistants either earn their keep or embarrass themselves, and a 7-point lead over GPT-5.4 in a single release cycle is more than cosmetic. Whether that translates to real autonomy on production codebases is the open question.

The one thing I'd watch: API access is delayed for cyber safeguard reasons, and OpenAI is shipping stricter classifiers for potential cyber risk which some users may find annoying initially. How tight those guardrails feel in practice will shape developer adoption more than any benchmark.

What do you think? Drop your thoughts in the comments.

FAQ

When was GPT-5.5 released?

GPT-5.5 was released on Thursday, April 23, 2026, rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex.

How much does the GPT-5.5 API cost?

The standard API is $5 per 1M input tokens and $30 per 1M output tokens. GPT-5.5 Pro costs $30 input and $180 output per 1M tokens.

What is GPT-5.5 best at?

Agentic coding, computer use, knowledge work, and early scientific research. It leads Terminal-Bench 2.0 at 82.7% and resolves 58.6% of SWE-Bench Pro tasks in one pass.

How does GPT-5.5 compare to Claude Opus 4.7?

On Terminal-Bench 2.0, GPT-5.5 scores 82.7% versus Opus 4.7's 69.4%. On OSWorld-Verified, GPT-5.5 scores 78.7% versus 78% for Opus 4.7.

Is GPT-5.5 available in the API right now?

Not yet. OpenAI says API access is coming soon, pending additional cybersecurity safeguards for large-scale deployment.