GPT-6 Astra for Business: Pricing, Admin Controls and OpenAI's Safety Numbers
OpenAI published a follow-up announcement on September 9, 2026 about GPT-6 Astra, the model it introduced the previous week. This post is the business and enterprise rollout rather than the model debut, and everything below comes from that announcement.
Transcript
OpenAI calls GPT six Astra its most capable model, built for the work businesses need to get done.
Astra is now available in ChatGPT Work, Codex and the A P I. It can operate the applications teams already use.
Pricing starts at ten dollars per million input tokens and fifty dollars per million output tokens. OpenAI says it completes tasks in fewer tokens.
On OpenAI's own safety test, Astra produced far fewer unintended outcomes than its previous flagship model. New admin controls limit what it can reach.
This is OpenAI's own announcement, with its own internal benchmarks. The model arrived last week, and enterprise access is off by default.
Inboxsmith builds AI phone and email agents for small business. Please like and subscribe for more news.
Sources
Every claim in this video comes from the top ranking coverage of this topic. The claims and where each one came from:
- Our most capable model, built for all the work businesses need to get done.(OpenAI's official announcement, "GPT-6 Astra: The next generation in intelligence for work", published September 9, 2026)
- Last week we introduced GPT-6 Astra, the world's most intelligent and aligned model, now available in ChatGPT Work, Codex, and the API.(OpenAI's official announcement, "GPT-6 Astra: The next generation in intelligence for work", published September 9, 2026)
- In ChatGPT Work and Codex, it can write code and work through the same applications people use every day, even when those applications don't have an API.(OpenAI's official announcement, "GPT-6 Astra: The next generation in intelligence for work", published September 9, 2026)
- Pricing starts at $10 per million input tokens and $50 per million output tokens.(OpenAI's official announcement, "GPT-6 Astra: The next generation in intelligence for work", published September 9, 2026)
- It's been trained to complete tasks in fewer tokens with fewer retries, which means less rework and lower cost per task.(OpenAI's official announcement, "GPT-6 Astra: The next generation in intelligence for work", published September 9, 2026)
- During training, we tested Astra on our internal computer use safety benchmark which tests models against the hardest business scenarios such as exposing confidential information, sharing a dashboard too broadly, or deleting data. In this evaluation, Astra produced unintended outcomes 89% less often than GPT-5.6 Sol and 74.7% less often than Claude Fable 5.1.(OpenAI's official announcement, "GPT-6 Astra: The next generation in intelligence for work", published September 9, 2026)
- New enterprise admin controls let them restrict access to approved websites and desktop applications, manage uploads and downloads, and control browsing history.(OpenAI's official announcement, "GPT-6 Astra: The next generation in intelligence for work", published September 9, 2026)
- Enterprise administrators can enable Astra under their applicable rate card and agreement. Enterprise access is off by default at launch.(OpenAI's official announcement, "GPT-6 Astra: The next generation in intelligence for work", published September 9, 2026)
- Honesty caveat: every claim in this video comes from a single source, OpenAI's own product announcement about its own model, and OpenAI is the party that benefits commercially from enterprises adopting Astra. The model itself was introduced the previous week, on September 3, 2026, so this post is the business and enterprise rollout rather than the debut of the model. The eighty nine percent figure is measured on an internal computer use safety benchmark that OpenAI designed, ran, and scored itself, and the announcement cites no independent audit, no third party replication, and no published methodology for that benchmark. The same applies to the efficiency and cost per task claims, which reference OpenAI's own positioning on evaluations such as Terminal-Bench 4.0 and the Artificial Analysis Intelligence Index. The favorable quotes in the announcement come from companies including Cognition, Databricks, Hebbia, Box, Figma, Thomson Reuters Labs, Datacurve, Basis, CodeRabbit, and XTX Markets, all of which build products on top of OpenAI models and therefore have a commercial interest in the model performing well. Pricing is stated as a starting price, not a full rate card, and enterprise administrators must enable Astra under their own agreement because enterprise access is off by default at launch. We captured zero video transcripts for this story, so no corroborating source was available at publication time, and readers should treat all performance numbers here as vendor reported until independent testing is available.(OpenAI's official announcement, "GPT-6 Astra: The next generation in intelligence for work", published September 9, 2026)