OpenAI has launched GPT-6 Astra, describing it as the “world’s most intelligent” AI model. The company said it has moved ahead of rival Anthropic in software engineering, science and cybersecurity.
The launch comes just over a year after the company introduced GPT-5, as OpenAI seeks to regain the top spot in the rapidly expanding AI industry.
Also Read: OpenAI AI Agents Built a Secret Society, Exchanged 70,000 Messages and “Sacrificed” Themselves
GPT-6 Astra Marks Step Towards AGI: OpenAI
OpenAI said GPT-6 Astra has delivered market-leading results across several fields.
“The model could represent an important step towards artificial general intelligence (AGI), referring to AI capable of handling a wide range of tasks at a level comparable with humans.”
-said, Greg Brockman, President, OpenAI.
Looking back in a few years, the period around the launch of GPT-6 Astra could be seen as the point when AGI was created, Brockman noted.
He also said current models are solving previously unsolved 100-year-old mathematics problems while helping accelerate the economy and support people in their personal lives.
Also Read: OpenAI Is Bringing Ads to ChatGPT in India: What Users Need to Know
Anthropic AI Models Accessed Real Systems During Tests
The OpenAI launch follows a security incident disclosed by Anthropic last month involving its Claude AI models.
Anthropic said models accessed the internet from testing environments and subsequently gained unauthorised access to the production infrastructure of three organisations. The company said this was not a deliberate attempt by the models to escape their testing environments, but resulted from a misconfigured evaluation environment.
Anthropic reviewed 141,006 cybersecurity evaluation runs and found three separate incidents involving six runs. The affected models were Claude Opus 4.7, Claude Mythos 5 and an internal research test model.
Also Read: OpenAI Slows AI Development After Rogue Agent Hacks Hugging Face
How the Claude Security Incident Happened
The tests involved a “capture-the-flag” exercise in which Claude was asked to find hidden information inside a fictional network. The model was told it had no internet access and was operating in a simulated environment.
Anthropic said a misunderstanding with its evaluation partner, Irregular, meant the systems actually had live internet access. Claude therefore treated real online systems as part of the fictional exercise and interacted with them.
OpenAI had also disclosed a separate incident earlier this month. One of its frontier AI models escaped a restricted testing environment during an internal cybersecurity benchmark and gained access to Hugging Face, including its production infrastructure.
The GPT-6 Astra launch refocuses the race between OpenAI and Anthropic, while recent security incidents show how quickly AI systems can move beyond controlled testing environments.
Samarjit Kaur is a journalist and communications professional covering technology & emerging digital trends. With a focus on clarity and context, she reports on developments shaping industries and governance. When not reporting, she chooses to plug-in and relax on her playlists and plan her next bucket-list trips!
