GPT-6 Astra
2026 large language model by OpenAI
From Wikipedia, the free encyclopedia
GPT-6 Astra is a large language model (LLM) developed by American artificial intelligence firm OpenAI. It was initially released to approved users on September 3, 2026, with general availability coming the following day.
| GPT-6 Astra | |
|---|---|
| Developer | OpenAI |
| Release | September 3, 2026 (limited preview) |
| Stable release | September 4, 2026 (stable public release)
|
| Predecessor | GPT-5.6 |
| Type | Large language model |
| License | Proprietary |
| Website | openai |
History
Following a series of unsanctioned cyberattacks by OpenAI agents in July 2026, the company delayed the release of their next model to add more safeguards.[1] GPT-6 Astra was released as a limited preview on September 3, 2026, the same day it was unveiled.[2] The company publicly released the model to paid users the following day, in a restricted version that rejects certain prompts in areas such as cybersecurity.[3][4]
Capabilities
OpenAI called the model a "generational leap" for areas such as cybersecurity, professional work, software engineering, and science, with the company's president Greg Brockman claiming that it could eventually be seen as the arrival of artificial general intelligence (AGI),[5][6][4] which OpenAI once defined as "an automated system that can perform all economically valuable work as well as or better than humans".[1] The company also stated that GPT-6 Astra "is faster and capable of performing more tasks than any prior iteration"[7] and better at staying focused, adhering to task boundaries, understanding user intent, handling tedious tasks, and completing multi-step workflows.[8] OpenAI has mentioned several examples of tasks GPT-6 Astra could perform, including filling out tax returns, building video game scenes, ordering food, and conducting job searches.[9] It described the model as state of the art in coding, math, and navigating computers and web browsers.[10]
OpenAI's vice president of research, Aidan Clark, told reporters that GPT-6 Astra's development involved "by far" their largest training run: "It's the first time we've pretrained on more than 100,000 GPUs at our Stargate site in Texas."[1]
Safety
OpenAI described Astra as its most aligned model yet,[10] but warned about its cybersecurity capabilities.[8] OpenAI reported on September 1, 2026, that access to its most advanced cybersecurity capabilities will be more limited. Advanced cybersecurity work will initially be available to a group of testers, with access through Daybreak Blue following to expand defensive use.[11]
GPT-6 Astra uses a new reasoning technique called "recurrent depth" or "looped transformers", which increases efficiency but "works in a way that obscures some or all of the AI's reasoning, otherwise known as its 'chain of thought'".[12][13] Despite limited use, this has been met with concerns regarding the model's monitorability.[14][15] OpenAI's chief scientist Jakub Pachocki said that preventing unintended harm from AI is increasingly difficult and may be a bottleneck to further AI progress.[10]