OpenAI launches GPT-6 Astra, Sam Altman calls it best AI model in the world

OpenAI has finally released GPT-6 Astra, its most powerful AI model yet. The AI startup claims Astra is so good that it may start the era of AGI, or artificial general intelligence.Astra is OpenAI's most powerful AI model yet. (Photo: Reuters) After weeks of teasers around Astra and what it could do, OpenAI has finally released the AI model. GPT-6 Astra, OpenAI says, is its most powerful AI model yet, and could be the beginning of AGI. Going by benchmarks shared by OpenAI, Astra is the most advanced AI model currently available, surpassing Anthropic’s Claude Fable 5.1.Announcing the release, OpenAI CEO Sam Altman wrote on X that GPT-6 Astra was the “best model in the world for computer use, professional work, science, coding, cybersecurity, and more.” Altman believes that the model will "enable a new generation of entrepreneurship, scientific discovery, and building." Sam Altman X post. In a media briefing, OpenAI president Greg Brockman said, "It’s not unreasonable to feel that we are now in the AGI era.” Brockman was confident that in the future, people may look back at Astra’s release as a pivotal moment in the world of AI. "I think that if we fast-forward a couple of years, when we look back and say, ‘When was it really that AGI was created?’ I think it's going to be about this time, and I think it might be about this model,” he explained.AGI, or artificial general intelligence, refers to a state where AI can think like humans, if not surpass human intelligence. Astra’s release comes just weeks after GPT-5.6 Sol, which was launched in July.How good is GPT-6 Astra?Astra goes a step further when it comes to tasks like coding. In a blog post, OpenAI called GPT-6 Astra the "best model for software engineering to date." The startup pointed to benchmark results showing that Astra outperformed OpenAI’s own GPT-5.6 Sol and Anthropic’s Fable models on tasks such as finding bugs, executing terminal tasks and answering questions about codebases. In terms of benchmarks, Astra outperforms virtually every other AI model till date. It scores 98 per cent on FrontierMath Tier 4, 99.9 per cent on ARC-AGI 3. On Terminal-Bench 4.0, OpenAI said Astra scored 57.9 per cent, compared with 37.3 per cent for GPT-5.6 Sol and 55.8 per cent for Claude Fable 5.1.In a launch video posted online, OpenAI focused on computer use capabilities of Astra. In the clip, we can see Astra completing different tasks based on requests from users, including creating 3D models in Blender and making presentations all by itself.OpenAI says that Astra can handle tasks such as filling online forms, organising calendars, conducting online research, or analysing scientific data. It can also create websites, install and test software. Astra can also be used to draft legal documents or prepare tax returns.GPT-6 Astra’s cybersecurity capabilitiesIn the past few months, there has been growing discussion over AI use in cybersecurity. After the launch of Anthropic’s Claude Mythos and OpenAI’s GPT-5.6 Sol, we had seen AI models that were getting better at finding cybersecurity flaws. Astra goes a step further in this area too.As per OpenAI, in cybersecurity testing without production safeguards, Astra scored 100 per cent on ExploitBench, compared with 78.5 per cent for GPT-5.6 Sol. OpenAI says that Astra’s ability to identify and develop zero-day exploits could help defenders find and patch weaknesses, but also required stronger safeguards.Though OpenAI is keeping safety as a key focus after a recent incident involving other OpenAI agents and Hugging Face’s systems, which raised fresh concern about alignment and monitoring. OpenAI claims Astra is its “most aligned mode yet.” The AI startup conducted an evaluation based on the Hugging Face incident with Astra. It found that GPT-6 Astra went beyond an authorised target in 0 per cent of cases, compared with 48% for GPT-5.6 Sol without production safeguards.The company states that Astra would refuse more advanced cybersecurity tasks such as creating proof-of-concept exploits for vulnerabilities at launch.Can you use GPT-6 Astra?As per OpenAI, GPT-6 Astra is rolling out to customers using Daybreak, OpenAI’s cybersecurity programme, and would over the next week become available to ChatGPT Plus, Pro, Business and Enterprise users. The model will also be available through the OpenAI API, AWS and Amazon Bedrock.Addressing the delay to ChatGPT users, OpenAI’s Codex head Thibault Sottiaux announced that the company was giving free limit resets to users till Astra was made available. “We will give one banked reset for every day you don't have access to Astra on your paid ChatGPT plan, starting today,” he wrote on X.In a follow-up post, Sam Altman apologised for the phased release of Astra. "First, sorry for the messy rollout. second, when we screw up, we try to make it right,” he wrote. “Third, we should be able to begin broad rollout to API customers and ChatGPT subscribers in the near future. As usual we will start with pro subscribers.” Sam Altman apologised for a phased release of Astra. The launch places Astra at the centre of OpenAI’s push to expand business use of its models. The AI startup is expected to go public with an IPO set for later this year.- EndsPublished By: Armaan AgarwalPublished On: Sep 4, 2026 08:39 IST