GPT-6 Astra Is “Here.” So Who Can Actually Use It?

Evidence boundary — October 14, 2026: The hosted ChatGPT plan, Work/Codex rollout, Enterprise/Edu and Daybreak access statements below are dated October 9–11 reporting, not verified current account terms or entitlement. Five official Help Center pages could not be retrieved substantively in this check; their transport and browser failures are retained. The new official API/model, usage-rate, data-control and cloud documentation captures support only the exact public facts identified below. Kingy made no paid Astra API request and did not repeat the historical evaluations.

Original access-report cutoff: October 11, 2026. The September updates preserve the original launch-period account. Supplemental public-document check: October 14, 2026, for the API/model table, published credit meter and identified cloud documentation. Hosted plan status and account-specific access remain unverified; a provider document does not confirm access for a specific account, project, region or quota.

Dated rollout log: The following September entries preserve what this article reported at the time. They do not certify current hosted-plan entitlement, early-access status, Azure numeric prices or a newly verified account result.

September 8, 9:05 p.m. Pacific update: AWS now confirms that GPT-6 Astra is generally available on Amazon Bedrock. Its new model card documents openai.gpt-6-astra on the Bedrock Mantle endpoint in Oregon, plus us.openai.gpt-6-astra and global.openai.gpt-6-astra cross-Region inference profiles on Bedrock Runtime. AWS also publishes Standard pricing and says Priority and Flex are not supported for Astra. The Bedrock row therefore moves from Reported to Confirmed.

September 7, 5:02 p.m. Pacific update: Microsoft’s current Foundry model catalog now labels gpt-6-astra generally available. Microsoft also says no separate model-access request is needed, although some Azure subscription tiers must still request quota and Tier 5 and Tier 6 subscriptions receive quota by default. The Azure row therefore moves from rollout-gated to generally available, while retaining the subscription, region, quota, and capacity caveats that determine whether a particular account can deploy it.

September 6, 12:57 p.m. Pacific update: OpenAI’s official rollout announcement says Astra is live in the API and available to all Pro, Enterprise, and Business Premium users in Work and Codex. OpenAI leaders then said the Work and Codex rollout had also reached all Plus and Business users. The access label for those surfaces therefore moves from phased rollout to rollout complete. This does not add GPT-6 Pro to ordinary Plus Chat, remove Enterprise and Edu administrator controls, or guarantee that every API project, Azure subscription, or region has usable entitlement and quota.

September 5, 8:51 p.m. Pacific update: OpenAI’s current Astra model guide adds a regional limit to the Fast option. API projects using EU data residency cannot use service_tier: "fast" or service_tier: "priority" with Astra and must use Standard processing. OpenAI also says Astra Fast has no latency service-level agreement.

September 5, 4:50 p.m. Pacific update: OpenAI’s current Enterprise and Edu guidance corrects an earlier two-week assumption. In eligible Enterprise and Edu workspaces, Astra is off by default and rolls out gradually; owners must enable it through model-access controls after it reaches the workspace. Existing Early Model Access settings do not carry over, and Astra will not automatically turn on after two weeks.

September 5, 12:49 p.m. Pacific update: Microsoft now publishes Astra-specific Azure quota limits. At Tier 1, gpt-6-astra receives 333 requests and 333,000 tokens per minute for Data Zone Standard, or 1,000 requests and 1 million tokens per minute for Global Standard. Foundry assigns and upgrades quota tiers at the subscription level, so these documented limits still do not guarantee that a new subscription has deployment access or usable capacity.

September 4, 8:48 p.m. Pacific update: OpenAI then published estimated Astra capacity for local Work and Codex messages in each five-hour period: 5–45 on Plus and Business Standard, 25–225 on Pro $100, and 100–900 on Pro $200. These are estimates, not fixed caps; local messages and cloud chats share an allowance, and weekly limits may apply. OpenAI also says that during the initial Enterprise rollout, an organization needs Daybreak access before an administrator can enable Astra.

September 4, 4:49 p.m. Pacific update: Microsoft’s Foundry model catalog now lists gpt-6-astra, version 2026-09-03, and links to a live region matrix. Microsoft says some Azure subscription tiers must request quota before deploying it; Tier 5 and Tier 6 subscriptions receive quota by default. This confirms the Azure model ID and deployment documentation, but not access for every subscription.

September 4, 12:45 p.m. Pacific update: OpenAI now says Astra requires Codex CLI 0.153.0 or newer and has published token-based Work and Codex credit rates: 250 credits per million input tokens, 25 per million cached input tokens, and 1,250 per million output tokens. Fast mode costs 2.5 times the Standard credit rate in Work and Codex, versus 2 times Standard in the OpenAI API.

Earlier September 4 update: OpenAI published separate Astra entitlement guidance for ChatGPT Work and Codex. Pro $100, Pro $200, and Business Premium can use their full existing Work and Codex allowance for Astra; Plus and Business Standard receive limited Astra usage, with optional credits after the included amount is used. Credits do not provide early rollout access.

Earlier September Chat plan update: OpenAI documented Astra as GPT-6 Pro in ChatGPT and published plan-specific Chat allowances. At that point, OpenAI described Chat access as rolling out to Pro $100, Pro $200, Business, and Enterprise; Plus was explicitly excluded from GPT-6 Pro in Chat.

Earlier update: Microsoft confirmed that GPT-6 Astra began rolling out through the Microsoft Foundry Limited Access Program and published the supported deployment types and Azure-specific token prices.

GPT-6 Astra is officially here. OpenAI now has a GPT-6 Astra system card, a developer model page, an API changelog entry, and ChatGPT release notes. That still does not mean Astra has reached every eligible account.

September 2026 launch-period wording: OpenAI’s ChatGPT release notes described access as rolling out to a limited set of organizations, said Astra was not generally available, and planned broader availability over the following days. That was the launch-period status, not a description of current availability. The original reporting contrasted broader Plus, Pro, Business and Enterprise launch wording with a specific plan guide excluding GPT-6 Pro from Plus in Chat. That hosted-plan comparison is retained as dated reporting; current hosted eligibility was not verified in this check. Product surface matters.

That gap matters because there are at least three materially different Astra experiences: the model announced to the press, the safeguarded production model intended for ordinary ChatGPT and API use, and the Daybreak Blue configuration behind Astra’s headline cybersecurity results. Those should not be treated as interchangeable.

GPT-6 Astra access matrix: dated hosted reporting and confirmed public API facts

Evidence labels: Hosted-plan rows preserve the October 9–11 report and do not confirm current account terms. “Confirmed public documentation” identifies exact facts in the new October 14 provider captures, not a successful paid request or universal entitlement. “Unverified” preserves a question the available evidence does not settle.

Pro plan reporting from October 9–11, 2026 — current hosted terms unverified: OpenAI lists Pro 100 at $100/month, Pro 200 at $200/month, and Pro 500 at $500/month; Pro 500 includes Astra Ultrafast. New Pro 200 subscriptions that do not qualify for grandfathering have a lower usage allowance. Eligible customers who had an active Pro 200 subscription from September 22 through September 29 at 10 a.m. Pacific can keep the previous allowance through October 29 while their subscription remains active. These plan details do not establish the exact GPT-6 Pro Chat allowance for an individual account. Check your account before changing plans. See our Pro vs Max buying guide before choosing a plan.

The three Astra experiences

1. The model announced to the press

This is the GPT-6 Astra in launch headlines: the new flagship model, its benchmark claims, and access across paid ChatGPT plans, the OpenAI API, AWS Bedrock, and Microsoft Azure. OpenAI now officially uses the GPT-6 Astra name and calls it its most capable model. The October 11 report described completion of the paid-plan rollout in Work and Codex. That hosted rollout is not independently verified in this check. Current official captures document the public API, AWS routes and the Azure model/version; hosted Chat and managed-workspace access, cloud quota and capacity remain separate questions.

Astra’s official product documentation is now live, but general availability remains product-specific. The original report described Microsoft Foundry as generally available; its current catalog capture confirms the model/version and quota guidance but does not settle that access claim. AWS documents Amazon Bedrock routes; neither provider-wide label guarantees that every account, region, inference option, or quota is ready. The practical reading is that announcement day, provider availability, account entitlement, and a successful production request can occur at different times.

2. The safeguarded production Astra

This is the version ordinary ChatGPT, Codex, and API customers should expect. OpenAI says its misalignment monitoring reviews an agent’s reasoning, actions, inputs, and outputs across ChatGPT, Codex, and supported Responses API use. A flagged conversation can be paused or ended. Some product surfaces let the user review and resume; conversations stopped through the API may not be resumable.

That difference is operational, not cosmetic. A long autonomous workflow needs to handle a stopped API response without silently retrying the same action forever. OpenAI also says this misalignment monitor does not apply to Chat Completions, because that interface cannot combine reasoning and tool use; separate model refusals and misuse safeguards still apply.

3. Astra evaluated with Daybreak Blue

OpenAI’s September 3 system card reports its headline Astra cyber results with Daybreak Blue, not the default production configuration. Daybreak guidance cited in the October 11 report distinguished Blue, with standard safeguards on Astra and GPT-6.1 Sol, from Red, with reduced safeguards; current program terms were not verified in this check. This distinction describes safeguards, not an account entitlement or a promise of reduced refusals for every Blue user.

The September 3, 2026 system card reports the gap. On Astra, its Daybreak Blue evaluation raises proof-of-concept exploit completion from 2.4% to 92% and cyber red-teaming completion from 7.4% to 76.9%. Even with Blue, Astra fully completes 3.5% of the arbitrary cyber requests in OpenAI’s Advanced Cybersecurity Completion Rate evaluation. These are dated safety-evaluation results, not general benchmarks of model intelligence or a claim that every current Daybreak account receives Blue.

The original article cited an FAQ mapping the API alias gpt-daybreak-blue-latest to gpt-5.6-sol and an AWS mapping of Daybreak Blue to openai.gpt-daybreak-blue-5.6-sol. Treat those alias names and mappings as dated source observations: the current source captures retained for this update do not verify them or establish an Astra Daybreak alias.

So a Daybreak badge, a Daybreak alias, and Daybreak-configured Astra are three different things. If you need the last one, ask OpenAI to confirm the underlying model, access level, and safeguard profile in writing.

API ID, pricing, limits, and data controls

Earlier rate-limit snapshot: The October 11 version of this article recorded Tier 1 at 500 RPM / 500,000 TPM, Tier 2 at 5,000 / 1 million, Tier 3 at 5,000 / 2 million, Tier 4 at 10,000 / 4 million, and Tier 5 at 15,000 / 40 million, with batch-queue limits from 1.5 million to 15 billion tokens. Those earlier tier labels and limits are retained as dated report history; the current model page uses the Build, Launch, and Grow table above.

At the official API Standard rates, a request that consumes 100,000 uncached input tokens and 20,000 output tokens would cost about $2: $1 for input and $1 for output. API Fast would make the same token mix about $4, before tool or service-specific charges. The long-context surcharge begins above 272,000 input tokens. Work and Codex use a separate credit meter: the mix consumes 50 Standard credits, 100 credits at the credit-billed Fast rate, or 125 credits’ worth of included allowance at the Fast allowance-consumption rate.

Fast and Ultrafast are service tiers, not higher-intelligence model variants. Their multipliers describe allowance consumption or credit billing, not a guaranteed speed increase. Kingy’s Fast mode pricing guide explains the speed-versus-cost tradeoff on the previous generation.

Historical Azure price report, October 11: This article recorded the following separate Foundry prices; those numeric Azure rates were not independently reconfirmed by the October 14 catalog capture. Standard Global short-context deployments match OpenAI’s $10 input and $50 output rates, with $1 cached input and $12.50 cache writes per million tokens. Long-context Global is $20 input, $2 cached input, $25 cache writes, and $75 output. U.S. Data Zone rates are $11/$1.10/$13.75/$55 for short context and $22/$2.20/$27.50/$82.50 for long context. Those are Azure deployment prices, not OpenAI’s Fast tier. The October 11 article also reported general availability without a separate model-access request; those access assertions remain unverified in this check. Current Foundry documentation assigns quota by subscription; its Tier 1 Astra limits are 333 RPM / 333,000 TPM for Data Zone Standard and 1,000 RPM / 1 million TPM for Global Standard. The direct OpenAI API is the easier test path when gpt-6-astra is already enabled in your OpenAI project and you do not need Azure governance or regional deployment controls.

AWS’s Bedrock Standard rates follow the same short-versus-long-context split. Global cross-Region inference costs $10 input, $12.50 cache writes, $1 cache reads, and $50 output per million tokens through 272,000 input tokens; above that threshold, the rates are $20/$25/$2/$75. In-Region and U.S. geographic cross-Region inference cost $11/$13.75/$1.10/$55 for short context and $22/$27.50/$2.20/$82.50 for long context. AWS’s current model card lists Ultrafast at six times the corresponding Standard prices and documents it on Bedrock Runtime through U.S. geographic and Global CRIS, and on Bedrock Mantle in us-east-1. These Ultrafast routes and prices differ from the existing Standard Mantle region. Priority and Flex are not supported for Astra. The October 11 article reported launch guidance allowing up to 30 days of classifier-flagged traffic retention and requests for Zero Data Retention through an AWS account team. Those retention/access terms were not independently reconfirmed in the October 14 model-card capture.

How to check whether you can use Astra

In ChatGPT

1. Open the model picker in the specific workspace you intend to use.
2. Look for the model name actually offered to your account. The October 11 report described Astra as GPT-6 Pro in ChatGPT; this check does not confirm current hosted-plan availability.
3. Confirm current eligibility in the model picker. The October 11 report excluded GPT-6 Pro from Plus in Chat and listed Pro 500/Ultrafast separately; current hosted eligibility was not verified here. Workspace controls and account-level limits still apply.
4. If you are in a managed workspace, ask an admin to check model controls and regional restrictions.
5. For cyber work, note whether the workspace is in Daybreak and whether Astra itself has the requested access level. A general Daybreak approval does not settle that question.

In Codex

1. The October 11 report specified Codex CLI version 0.153.0 or newer; check the current client requirements and model picker rather than treating that historical minimum as current guidance. The report described the paid-plan rollout in Work and Codex as complete, but access still differs from Chat and can depend on the signed-in account, seat, workspace controls, and app version.
2. If you sign in with ChatGPT, check the current plan and account usage guidance before budgeting Astra. The October 9–11 report described a Pro $200 allowance change after the September estimates; current hosted allowance terms were not verified here. Weekly limits may apply. If you use your own API key, API pricing applies instead.
3. Do not buy credits expecting access that your account does not show. The October 11 report said credits did not grant earlier rollout access; a published credit rate does not establish current entitlement.
4. If Astra is absent, do not substitute gpt-daybreak-blue-latest and call the result Astra. An earlier FAQ mapped that alias to GPT-5.6 Sol; the current source captures do not verify its target. Confirm the returned model ID or ask OpenAI to confirm the underlying model and safeguard profile.
5. Plan for monitored cyber tasks to pause for review. That is a stated product behavior, not proof that your account already has Astra.

If Astra has not reached your account, use Kingy’s OpenAI model selection guide to choose the best documented model available now.

In the OpenAI API

1. Confirm that gpt-6-astra appears in the model list available to your project.
2. Use the Responses API for tool calling. Although the model also supports Chat Completions, OpenAI’s changelog says Astra tool calling requires Responses.
3. Make a minimal request, then record the UTC time, project, endpoint, returned model value, and request ID.
4. If the project uses EU data residency, use Standard processing; OpenAI’s current Astra model guide says API Fast is unavailable with EU residency. Check the current project documentation for the supported service tiers.
5. Test the failure path. A monitored task can stop rather than opening an interactive review step.
6. Check your organization’s limits page before moving production traffic.

Use OpenAI’s official API quickstart and API changelog for the request shape and migration constraints. A copied launch-day code sample cannot grant project entitlement.

For Daybreak Blue

The October 11 report described Daybreak application, identity/trust verification and approval, plus Enterprise administrator enablement and a destination project key. Current onboarding requirements were not verified in this check; confirm them with the provider before proceeding. The public sources checked here do not establish a per-member Advanced Account Security prerequisite for Daybreak Blue. Daybreak and Zero Data Retention have separate approval paths.

The official enterprise Daybreak onboarding guide is the right checklist. The important Astra-specific question for your account team is: “Which underlying model, Daybreak tier and safeguard profile are enabled for this project?”

What is confirmed, and what is still moving

OpenAI’s September 3 system card reports GPT-6 Astra’s Critical cyber-capability threshold and that the headline cyber evaluation used Daybreak Blue rather than the default production setup. It also confirms that external monitoring covers ChatGPT, Codex, and supported Responses API use, while a stopped API conversation may not be resumable.

OpenAI’s September distribution updates described Astra as GPT-6 Pro in Chat; listed Pro $100, Pro $200, Business, and Enterprise as eligible as rollout reached them; and excluded Plus from GPT-6 Pro in Chat. The original report also cited an API release and completion of the Work/Codex paid-plan rollout. The current API release is supported by the retained official changelog; the hosted rollout claim remains dated reporting. The September five-hour local-message estimates preceded the Pro $200 allowance revision; check current plan and account guidance for capacity. Local and cloud work share plan usage, and weekly limits may apply. During the initial Enterprise rollout, OpenAI said an organization needed Daybreak access. OpenAI’s initial Enterprise and Edu guidance described gradual workspace rollout with the model off by default; owners had to enable it, existing Early Model Access settings did not carry over, and it would not switch on automatically after two weeks. The October 11 report specified Codex CLI 0.153.0 or newer; current minimum client requirements were not reverified. At the published Work and Codex rate card checked October 14, Fast uses 2.5× Standard against included usage and 2× the Standard credit rate; Ultrafast uses 8× included usage and 6× the Standard credit rate. The API model ID is gpt-6-astra; Standard API pricing remains $10 input and $50 output per million tokens, with API Fast at twice those rates. Astra Fast is unavailable for API projects using EU data residency and carries no latency SLA. The original report described the same model ID as generally available in Foundry and recorded Azure-specific prices. The current catalog/quota captures support its exact model/version, region references, subscription-tier rules and Tier 1 limits, but do not independently confirm those price or general-access assertions. AWS’s current model card documents the openai.gpt-6-astra model ID and cross-Region profile IDs, and publishes supported endpoints, regions, Standard prices and quota behavior. The earlier AWS retention/ZDR guidance remains dated article reporting, not a newly verified policy.

Official product pages still need to settle the remaining access questions: account-level exceptions after the announced Work and Codex rollout, exact Chat and managed-workspace dates, Healthcare eligibility, and which AWS accounts and Azure subscriptions have usable quota and capacity. Kingy did not make a paid Astra API request for this review, so this page cannot verify project-level access or performance.

For anyone tracking the launch from rumor to documentation, Kingy’s earlier Astra evidence tracker remains the paper trail. The model should appear in your product or project, the provider should document the exact configuration, and a request should succeed under the limits you intend to run. Treat your project as unprovisioned until all three are true.

The Kingy Brief

Get The Kingy Brief.

AI changes, original tests and one practical thing to try. Fridays at 09:00 Vancouver time.

Free · Double opt-in · Unsubscribe anytime

Privacy policy

Sign up for The Kingy Brief on Beehiiv.

Signup help and newsletter schedule

Signup form provided by Beehiiv. After submitting, check your inbox for "Confirm your subscription to The Kingy Brief" and open its confirmation link. Check Spam or Promotions if you cannot find it.

Fridays at 09:00 Vancouver time: source-checked AI changes, original tests and one practical thing to try. We skip a week when there is not enough verified material. Free. Unsubscribe anytime.

Sign up free