OpenAI has launched GPT-6 Astra, its latest frontier AI model trained on over 100,000 GPUs at its Stargate facility in Texas. According to OpenAI, Astra marks a significant jump over its predecessor GPT-5.6 Sol, scoring high marks across various benchmarks including 97.6% on FrontierMath, 74.1% on DeepSWE coding, and 72.6% on the OSWorld 2.0 computer-use benchmark. Company executives suggested the system approaches their definition of general intelligence.
The model is rolling out initially to select enterprise users via the Daybreak program, with broader access coming to ChatGPT subscription tiers, API users, and cloud partners like AWS and Microsoft Azure. Standard API pricing is set at $10 per million input tokens and $50 per million output tokens, with a faster tier available at double the price. Alongside the model, OpenAI unveiled updates to its Codex environment, including persistent note-taking features across context windows.
Why it matters
Sets new technical benchmarks across reasoning, cybersecurity, computer operation, and scientific problem-solving.
Expands computer-use agent capabilities, allowing automated execution of complex digital workflows across standard software.
Source: the-decoder.com
