This is the companion to the GPT-6 Astra pricing front-page briefing. OpenAI’s launch post and the gpt-6-astra model page are the primary sheets. Alberta buyers typing “GPT-6 Astra pricing Canada,” “private AI security Alberta,” or “AI automation Alberta agents” need the operator traps, not a lab recap.
What the launch and model page list
OpenAI’s launch post says GPT-6 Astra is rolling out to a limited set of organizations first, and over the coming days will become available to ChatGPT Plus, Pro, Business, and Enterprise users, plus the OpenAI API, Microsoft Azure, and AWS Bedrock. It says Astra is included in subscription allowances, with credits for more. Pro, Business, and Enterprise also get GPT-6 Astra Pro. Enterprise requires an admin to enable it, and it is OFF by default. This desk is not inventing Plus exclusions — the launch page names Plus.
“GPT‑6 Astra is rolling out today to a limited set of organizations and over the coming days will become available to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as through the OpenAI API, Microsoft Azure, and AWS Bedrock.”OpenAI, GPT-6 Astra launch post, September 2026
The model page prices text tokens at $10.00 input, $1.00 cached input, $12.50 cache writes, and $50.00 output per million. Prompts with more than 272K input tokens are priced at 2x input and cache rates and 1.5x output for the full request. Fast mode is priced at 2x the applicable rates. Computer use, MCP, and hosted shell are supported tools on the Responses API path.
| Line | Per 1M | Note | |
|---|---|---|---|
| Input | $10.00 | Standard | |
| Cached input | $1.00 | Prompt cache | |
| Cache writes | $12.50 | 1.25x uncached input | |
| Output | $50.00 | Standard |
Critical cyber and monitor pauses
OpenAI’s Path to Astra (Sep 1) says Astra is the first model designated at the Critical cybersecurity capability threshold under its Preparedness Framework. Advanced cybersecurity access is initially more limited, with Daybreak Blue expansion for defensive use. On OpenAI’s cyber jailbreak evaluation set, Astra refuses 91.5% of requests versus 59% for GPT-5.6 Sol — that is OpenAI’s number on their set.
Path to Astra and the system card hub also say misalignment monitoring can slow, pause, or stop legitimate long agent work. In ChatGPT or Codex, a user may be asked to review before continuing; in the API, the task stops. The system card hub states CoT monitorability decreased versus Sol. Treat those as OpenAI’s claims. For AI automation Alberta agents, design a human resume path and prefer action and tool logs over trusting chain-of-thought alone.
What Alberta operators should do
Price is not residency. A $10 / $50 USD line does not put inference in Alberta. For private AI security, do not paste tenders or credentials into computer-use sessions. Leave Enterprise off until a named owner enables it and scopes computer-use. Pin reasoning.effort; do not leave max on every call.
OpenAI’s Aug 28 Cursor decision still proposes a Nov 12 shutoff and says future models will not be provided to Cursor under that notice. Do not write Astra-in-Cursor as available. Keep a second model path. See the Cursor wind-down briefing.
FrontierMath Tier 4 98%, ARC-AGI-3 99.9%, and ExploitBench 100% are OpenAI’s published numbers from the launch post — not your shop eval. For AI consulting Alberta and AI automation Alberta work, run your own task suite before you promote a production agent.
The guardrail
Advanced cyber is not the default product. Daybreak / limited access is the rule on the Path to Astra note. Monitor pauses are a design input, not a surprise. Keep the pricing lead for the headline rates; use this sheet when you wire the agent.