Blog Intelligence
The $6-an-Hour Engineer? GPT-6 Astra, Explained
What OpenAI shipped
On September 3, 2026, OpenAI began rolling out GPT-6 Astra as a limited preview and called it a generational leap for software engineering, cybersecurity, science and professional work. OpenAI describes it as its best model for software engineering to date, and says ChatGPT is now close to twice as fast at computer use than before. President Greg Brockman went further, saying it is not unreasonable to feel we are now in the AGI era.
Where the “six dollars an hour” comes from
That number is not OpenAI’s. The API is billed per token: 10 dollars per million input tokens and 50 dollars per million output tokens, which is 2.5 times the promotional rate of the previous GPT-5.6 Sol model. The hourly framing comes from the newsletter Latent Space, which ran Astra as an autonomous engineer for two days on roughly 20 billion tokens and worked out the cost at under six dollars an hour. It is a useful thought experiment, not a plan you can subscribe to. Your real bill depends entirely on how many tokens your work burns.
What is actually new
The headline is not the price, it is the endurance. OpenAI is pitching Astra at long-session, agentic work: it drives a computer, browses, generates documents and slides, and does defensive security tasks like secure code review and patching. In Codex, OpenAI reports the updated harness finishes tasks about 1.9 times faster than the Sol experience. Astra is available across the paid ChatGPT tiers, the OpenAI API, and Amazon Web Services.
Why it matters for the stack
An hourly cost for a model is a category error that happens to be a useful one. It reframes a model as labor, priced against a contractor rather than against a rival API. The catch is that token bills are variable and open-ended in a way a salary is not. Astra is a real step up in autonomous engineering, but treat the six-dollar figure as one person’s benchmark run, not a quote.
Sources: for this explainer we checked OpenAI’s GPT-6 Astra announcement and Latent Space’s hands-on writeup where the hourly figure originates.
Common questions
- Did OpenAI price Astra at six dollars an hour?
- No. OpenAI charges per token: 10 dollars per million input tokens and 50 dollars per million output tokens. The hourly figure comes from an independent test, not OpenAI's price sheet.
- Where can I use GPT-6 Astra?
- In ChatGPT Plus, Pro, Business and Enterprise, through the OpenAI API, and on Amazon Web Services. It launched September 3, 2026 as a limited preview.
- Is it really better at coding?
- OpenAI calls it its best software engineering model to date and says its updated Codex harness completes tasks about 1.9 times faster than the previous GPT-5.6 Sol setup.