Safety overview: GPT-6 Astra
OpenAI's own safety overview: first model to reach Critical cybersecurity capability under the Preparedness Framework.
AI Storyline
5 items · 2 sources · 2 days
Operational story trace
GPT-6 Astra
Follow in this browser to see new updates on your Live feed.
Current stateLaunched · Critical cybersecurity tier
Latest change
A day-two recap frames Astra as OpenAI's biggest LLM launch yet: new SOTA computer-use and coding scores, 2.5x pricier per token but cheaper per completed task, and — flagged as a real caveat — reduced monitorability versus prior models.
Earlier contextThe story so far
OpenAI shipped GPT-6 Astra on September 3 as its most capable broadly deployed model, launch-day case studies showing agentic coding and document-review gains at Playco and Legora. It's also the first OpenAI model to hit the Critical tier for cybersecurity capability under the company's own Preparedness Framework.
editor-curated · source-linked
Arc
Sep 3Sep 4 · now
LAUNCH · Sep 3GPT-6 Astra ships as OpenAI's most capable deployed model, first to hit the Critical cybersecurity tier1 source · show source ▾
EARLY ADOPTION · Sep 3Playco and Legora case studies show agentic coding and document-review gains2 sources · show sources ▾
Playco cut manual fixes 50% prototyping games with GPT-6 Astra
Playco case study: 50% fewer manual fixes prototyping games with Astra.
Legora reviewed 41 documents in minutes with GPT-6 Astra
Legora case study: reviewed 41 financial documents in minutes, catching all four planted errors.
NOW · Sep 3–4Independent coverage frames Astra as cheap-per-task but less monitorable2 sources · show sources ▾
GPT-6 Astra: an automated AI Engineer you can hire for <$6 an hour
Latent Space frames Astra's economics as hiring an AI engineer for under $6/hour.
[AINews] GPT-6 Astra: OpenAI’s biggest LLM launch of all time
AINews recap: new SOTA computer-use and coding scores, 2.5x pricier per token but cheaper per completed task, and reduced monitorability.
Day 1 Thursday, Sep 3, 2026
Safety overview: GPT-6 Astra
OpenAI's own safety overview: first model to reach Critical cybersecurity capability under the Preparedness Framework.
Playco cut manual fixes 50% prototyping games with GPT-6 Astra
Playco case study: 50% fewer manual fixes prototyping games with Astra.
Legora reviewed 41 documents in minutes with GPT-6 Astra
Legora case study: reviewed 41 financial documents in minutes, catching all four planted errors.
GPT-6 Astra: an automated AI Engineer you can hire for <$6 an hour
Latent Space frames Astra's economics as hiring an AI engineer for under $6/hour.
Day 2 Friday, Sep 4, 2026
[AINews] GPT-6 Astra: OpenAI’s biggest LLM launch of all time
AINews recap: new SOTA computer-use and coding scores, 2.5x pricier per token but cheaper per completed task, and reduced monitorability.
What to watch — open questions
- Does the Critical cybersecurity tier rating trigger any additional usage restrictions or safeguards for agentic deployments?
- What specifically makes Astra less monitorable than prior models, and does that matter for audit or compliance requirements?
- Do the 2.5x per-token price and lower per-task cost hold up outside OpenAI's own case studies?
How this thread was built
editor wrote the arc · 3 beatswatcher 1 status change
Storylines are threaded mechanically from the feed: stories that share a distinctive anchor across multiple days and sources. Each item links to its original source. The evidence trace, current state, and open questions are written by the editor routine and refreshed whenever a new beat lands.