OpenAI Releases GPT-5.5: Chatgpt's New Capabilities with 'Agentic & Omnimodal' Intelligence

Liputan6.com, Jakarta - OpenAI once recently made a launch of its latest flagship model, GPT-5.5, on April 23, 2026.

Known internally by the codename "Spud," this model is touted as a significant milestone towards a fundamentally new era of computing.

This next-generation model is not simply an update, but rather a "new level of intelligence" specifically designed to handle complex real-world tasks.

With numerous significant improvements, GPT-5.5 is expected to revolutionize the way users interact with AI, making it a smarter and more efficient assistant.

OpenAI ensures that this model will provide a superior experience, with a focus on enhancing agentic capabilities and omnimodal capabilities.

GPT-5.5 comes with a suite of features and capabilities designed to improve efficiency and performance.

Despite being a larger and more powerful model, OpenAI has managed to keep its per-token latency on par with GPT-5.4, thanks to a complete redesign of the inference system.

One of the most notable improvements is a context window of up to 1 million token.

Furthermore, GPT-5.5 is designed with enhanced agentic intelligence, enabling it to write code, research online, analyze data, create documents, and switch between tools more effectively.

These capabilities allow GPT-5.5 to grasp tasks more quickly, require less guidance, and check its own work until completion.

The latest model is also natively omnimodal, meaning it can process text, images, audio, and video in a single integrated system, opening up opportunities for richer and more flexible interactions.

For high-stakes reasoning, chatgpt's "GPT-5.5 Thinking" mode provides smarter and more concise answers by giving the model more internal computation time to verify its own assumptions before responding.

OpenAI has released benchmark results demonstrating GPT-5.5's superior performance across various domains.

In GDPval, the model achieved or exceeded expert-level performance on 84.9% of tasks across 44 professional knowledge domains, up from 83.0% achieved by GPT-5.4.

In Terminal-Bench 2.0, GPT-5.5 scored 82.7% in a complex command-line workflow benchmark, far surpassing Claude Opus 4.7's 69.4%.

For AI's ability to operate independently on a computer, OSWorld-Verified recorded a success rate of 78.7% for GPT-5.5.

In CyberGym, the model scored 81.8%, up from 79.0% for GPT-5.4 and 73.1% for Claude Opus 4.7.

OpenAI's internal Capture-the-Flag (CTF) challenge also demonstrated GPT-5.5's dominance, with a score of 88.1%, compared to 83.7% for GPT-5.4.

GPT-5.5's coding capabilities are also commendable, with a 58.6% score in the GitHub issue resolution evaluation on SWE-Bench Pro and a 73.1% score for a coding task with an estimated completion time of 20 hours on OpenAI's internal Expert-SWE.

In bioinformatics, BixBench scored 80.5% and GeneBench achieved 25.0%, both surpassing GPT-5.4.

Despite offering substantial performance improvements, the API pricing for GPT-5.5 has been adjusted.

For GPT-5.5, the API price is $5 per million input tokens and $30 per million output tokens, double that of GPT-5.4.

Meanwhile, GPT-5.5 Pro has an API price of $30 per million input tokens and $180 per million output tokens.

OpenAI explains that despite the higher price per token, GPT-5.5 requires fewer tokens to complete the same task, so the overall cost may not increase substantially.

A 50% discount is also available for bulk processing and flexible pricing, with priority processing charged at 2.5 times the standard rate.

In terms of security, OpenAI classifies GPT-5.5's cyber and biological capabilities as "High" under its Emergency Readiness Framework.

OpenAI also offers special access through its "Trusted Access for Cyber" program for verified security professionals.