GPT-6 Astra Takes AI Computer Use to the Next Level: What It Can Do

OpenAI’s latest model delivers stronger benchmark scores and faster performance on complex computer-based tasks

OpenAI has unveiled GPT-6 Astra, a new AI model focused on computer use, with the company highlighting improvements in speed, accuracy and safety. The model is designed to handle tasks that normally require a person to work directly with websites, software and digital tools.

From filling out online forms and updating customer records to organising calendars and conducting online research, GPT-6 Astra is built to perform a wide range of tasks with less manual intervention.

GPT-6 Astra Scores Higher on Agents’ Last Exam

One of the key benchmarks highlighted by OpenAI is Agents’ Last Exam, which evaluates AI agents on complex professional tasks performed in real software environments.

GPT-6 Astra achieved a score of 59.3% in the comparison, ahead of Claude Opus 5 at 55.5% and GPT-5.6 Sol at 53.6%.

The benchmark covers professional workflows across areas such as financial modelling, engineering and media production.

OpenAI also reports an efficiency advantage for Astra. At the highest-scoring settings in the comparison, the model used approximately 65% fewer output tokens than Claude Opus 5.

Faster Performance on OSWorld 2.0

Speed is another area where GPT-6 Astra showed an improvement.

In latency simulations on OSWorld 2.0, Astra achieved a score of 72.6%, compared with 65.7% for GPT-5.6 Sol.

The difference was also visible in the estimated time required to complete tasks. Astra took roughly 40 minutes per task, while GPT-5.6 Sol took approximately 75 minutes.

According to OpenAI, this translates into about a 47% reduction in task completion time compared with GPT-5.6 Sol in these simulations.

What Can GPT-6 Astra Do?

GPT-6 Astra is designed to work directly with computer interfaces and software. Its capabilities include filling out forms, updating CRM records, organising calendars and carrying out online research.

The model can also draft summaries in email or document editors, analyse scientific data, generate plots and create websites.

For developers and technical teams, Astra can perform frontend quality-assurance checks, install and test software, and help troubleshoot problems displayed on screen.

Applications Across Different Industries

OpenAI has demonstrated GPT-6 Astra across several areas, including game development, electrical engineering and everyday knowledge work.

The demonstrations show how computer-using AI can go o beyond generating text or code. Instead, the model can interact with software and carry out multiple steps required to complete a task.

This could make AI agents increasingly useful for workflows that involve several applications or require repeated actions.

A New Direction for AI Agents

GPT-6 Astra reflects a broader shift in artificial intelligence toward more autonomous computer interaction. Rather than simply explaining how something should be done, an AI agent can increasingly operate digital tools and complete the workflow itself.

The benchmark results presented by OpenAI indicate that Astra combines higher task accuracy with faster execution in the tested environments.

As computer-use technology continues to develop, models such as GPT-6 Astra could play a larger role in professional software workflows, research, development and routine digital tasks.