GPT-6.2 Astra

Pre-release · OpenAI gray-test signals vs public API status. Follow release and GPTProto access here — this page has no active gpt-6.2-astra route yet.

MODEL FAMILYOpenAIAI text generation

PUBLIC RELEASEUnconfirmedNo verified launch date

GPTPROTO ACCESSNot availableNo active route on this page

API PRICINGNot publishedRates pending verification

API MODEL IDgpt-6.2-astraReported in Codex; not a selectable public route

Release news and API access are separate milestones. This page will be updated when a verified GPTProto route is available.

View status

ABOUT THE MODEL

GPT-6.2 Astra API

A place to check what is known, what is still unconfirmed, and which text APIs you can evaluate today.

Status (as of Nov 4, 2026). GPT-6.2 Astra is not a publicly released OpenAI API model. OpenAI’s model catalog and API changelog still center on GPT-6 Astra (gpt-6-astra) as the flagship for the hardest end-to-end work. The identifier gpt-6.2-astra has appeared in some Codex usage breakdowns and in at least one shared response.completed payload since about Oct 31–Nov 1, while the product picker continued to show GPT-6 Astra. That pattern is consistent with a limited gray test or traffic split—not with a documented developer launch.

What remains open. OpenAI has not published a release date, list price, context window, system card, or selectable public route for gpt-6.2-astra. Explicit CLI/API selection outside the reported test cohort has returned model_not_found. Separate rumors around gpt-6.2-betelgeuse (“Bel”), brief unannounced API IDs such as rc-ada-or-vm / rc-ada-sh-vm, and unnamed “Internal Model” benchmark charts should be tracked separately until OpenAI ties them to a named public model. Media reports that GPT-6.1 Astra was shelved over safety concerns in late September explain why people expect a 6.2 jump—but “6.2 equals a fixed 6.1” remains speculation.

What to do on GPTProto today. This pre-release page has no active GPT-6.2 Astra route. A GPTProto API key does not unlock gpt-6.2-astra before the route is verified. Use a listed model such as GPT-6 Astra, GPT-6.1 Sol, or GPT-6.1 Terra, keep the model ID in one config value, and re-check this page when a verified public ID, rate, and GPTProto route exist.

Track the gray rollout

Codex usage charts and response events have shown gpt-6.2-astra while the picker still says GPT-6 Astra. Treat that as a reported silent split. Watch OpenAI’s changelog and model catalog for a public confirmation.

Separate rumors from the public API

A usage label is not a documented developer route. Betelgeuse/Bel, parameter-count claims, discount rumors, and “shipping this week” posts are unconfirmed. Do not copy GPT-6 Astra’s price sheet onto 6.2.

Check GPTProto access

GPTProto is not an active GPT-6.2 Astra provider on this page. A usable route needs a confirmed public model ID, request format, and current rate. Those details appear here only after verification.

Evaluate a listed model today

For work you can ship now, compare GPT-6 Astra (hardest end-to-end), GPT-6.1 Sol (near-Astra at lower list rates), and GPT-6.1 Terra (visual / computer-use / writing). Keep the model ID configurable for a one-line switch later.

REPORTED SIGNALS

Evidence vs what it does not prove

Keep Codex sightings and media rumors separate from a documented public launch.

Signal	Observed	Does not prove
Codex usage row gpt-6.2-astra	Multiple accounts posted Oct 31–Nov 1 usage charts with a separate gpt-6.2-astra line while the picker still showed GPT-6 Astra. Single-account shares in screenshots ranged roughly 1.6%–19.8% of that account’s window—not a global penetration rate.	A public launch, selectable Codex model, published price, or that every account is in the cohort.
response.completed model field	At least one shared response.completed fragment named model gpt-6.2-astra.	That you can select the ID yourself, or that OpenAI has documented the route.
Forced CLI/API select	Third-party checks reported codex --model gpt-6.2-astra returning 400 model_not_found outside the test cohort; the official gpt-6.2-astra model page has returned 404.	Nothing contradictory—this reinforces that public selection is not available.
Betelgeuse / Bel	Separate Discord/internal-tracker screenshots and X posts named gpt-6.2-betelgeuse or Bel around Oct 28–30.	That Bel and gpt-6.2-astra are the same model. Keep them in different buckets until OpenAI says otherwise.
Nov 4 RC / dual-VM flash reports	Media reported brief API list sightings of rc-ada-or-vm and rc-ada-sh-vm, rapid removal, 503s on early calls, plus an OpenAI engineering post chart with an unnamed Internal Model. Not confirmed by OpenAI as gpt-6.2-astra.	Official GPT-6.2 Astra specs, pricing, or that those RC IDs are the public product name.

AVAILABILITY

Release & API status

Source snapshot · Nov 4, 2026

Three checks before you build with a new model.

Official model release

A confirmed public announcement and release date.

Awaiting confirmation

Public API access

Published API documentation and an identifiable model route.

Awaiting confirmation

Available on GPTProto

A verified route with current pricing and working request details.

Not available

Model ID, pricing, and request examples will be added after verification.

PUBLIC LAUNCH

How to tell it is really live

A usage-log line alone is not a public release. Use this checklist before changing production routes.

01OpenAI API changelog names gpt-6.2-astra.
02The official model page and catalog list the ID (no 404).
03GET /v1/models with your own key returns gpt-6.2-astra.
04Codex lists it as selectable and codex -m gpt-6.2-astra no longer returns model_not_found.
05Deployment Safety Hub publishes a matching System Card or addendum.
06A usage-log line alone is not enough—wait for the items above before changing production.

BUILD WHILE YOU WAIT

Text models you can explore now

All text models

Start with a current model page, then check its live rate and supported inputs.

Listed on GPTProto

OPENAI

GPT 6 Astra

Explore an existing OpenAI API route while the next generation is being confirmed.

Text generation

View model & pricing

Listed on GPTProto

OPENAI

GPT 6.1 Sol

Review another current OpenAI route for workflows with different output requirements.

Text generation

View model & pricing

Listed on GPTProto

OPENAI

GPT 6.1 Terra

Evaluate a current model from another family alongside the OpenAI options.

Text generation

View model & pricing

Availability and rates can change. Check each model page before sending a request.

COMMON QUESTIONS

GPT-6.2 Astra API FAQ

Is GPT-6.2 Astra already live?

Gray-test signals exist; a public product launch does not. Some Codex accounts show gpt-6.2-astra in usage analytics and response events while the picker still says GPT-6 Astra. OpenAI has not listed the model in its API catalog, changelog, or as a selectable public route as of the Nov 4, 2026 snapshot on this page.

Is GPT-6.2 Astra available through GPTProto?

No. This pre-release page does not provide an active GPT-6.2 Astra route. A GPTProto API key does not unlock the model before that route is verified. A model ID, rate, and request example will be added here after access is confirmed. Explore current OpenAI models above.

When will GPT-6.2 Astra be released publicly?

OpenAI has announced no public release date. Social posts predicting “Monday,” “this week,” or “today” have already missed. Check OpenAI’s changelog and model documentation; a public release date and API availability may still differ.

How much will the GPT-6.2 Astra API cost?

No verified GPT-6.2 Astra rate is shown here. Do not copy GPT-6 Astra’s official Standard list prices ($10 input / $50 output per 1M tokens) onto 6.2. Gray-test usage does not establish billing. Pricing will be published when a supported route and its billing details are confirmed.

How do developers get GPT-6.2 Astra API access and an API key on GPTProto?

Not yet. GPTProto is not an active GPT-6.2 Astra API provider on this page, and there is no separate GPT-6.2 Astra API key. A GPTProto API key works with listed OpenAI models today, including GPT-6 Astra, GPT-6.1 Sol, and GPT-6.1 Terra.

Can I call gpt-6.2-astra directly today?

Not as a documented public API. Outside the reported Codex test cohort, explicit calls with model gpt-6.2-astra have returned model_not_found. Until OpenAI publishes the route, use a listed model and keep your integration ready to swap the ID.

Is GPT-6.2 Astra the same as Betelgeuse / Bel?

No public evidence confirms that. gpt-6.2-astra appears mainly in Codex usage reports; gpt-6.2-betelgeuse / Bel appears in separate tracker screenshots and posts. This page tracks them as distinct rumors until OpenAI connects them.

Should I wait for GPT-6.2 Astra or use a current model?

If you need a testable API today, use GPT-6 Astra for the hardest end-to-end work, GPT-6.1 Sol when cost matters more, or GPT-6.1 Terra for visual reasoning, computer use, and everyday writing. Keep the model ID in one config value so you can switch when a verified public GPT-6.2 Astra route lands.

Release news & model comparisons

Guides and comparisons for listed OpenAI models while gpt-6.2-astra stays unconfirmed.

All articles

What Is GPT-6 Astra? Features, Pricing, AGI Claims, and Work Use Cases

What is GPT-6 Astra? Explore its release, pricing, benchmarks, coding, computer-use and agent features while 6.2 remains pre-release.

GPT-6 Sol vs GPT-6 Astra: Which AI Agent Model Is Worth the Cost?

A production-focused Sol vs Astra comparison covering API list prices, reasoning, tool calling, and computer use.

GPT-6 Astra vs Claude Fable 5.1: Which Is Better in 2026?

Compare GPT-6 Astra vs Claude Fable 5.1 for coding, agents, benchmarks, API pricing, and cache costs.