OpenAI o: Release Date, Personal Agent Rumors, and API Status

What is OpenAI “o”?

In the reports circulating before DevDay, o is the name of a proposed personal agent rather than a normal chat model. The concept is: give the assistant a goal, let it plan and execute through cloud services for a long time, and return with progress or a finished result. That is different from a chat turn that stops when the response is generated.

The rumored comparison set includes Meta Muse and Grok Bot-style assistants: systems that turn a broad request into a continuing task with tools, memory and background execution. The screenshot and secondary coverage should therefore be read as a product concept report, not as evidence that Meta built or owns OpenAI's product. No OpenAI page currently establishes the final name, user interface or launch scope.

OpenAI o release date: why September 29 matters

September 29, 2026 is the expected OpenAI DevDay date in San Francisco. OpenAI Developers posted a “72 hours” countdown on September 27 with the message that the team had been building and was ready to show its work. That post is an official event signal, not an “o” launch announcement.

Evidence	What it establishes	What it does not establish
OpenAI Developers countdown	DevDay was approaching and OpenAI planned announcements or demonstrations	That o is confirmed, or that every rumor will appear
Pre-event reports and code/document traces	A product concept named o may be under discussion or development	A final product name, model, price, API or ship date
September 29 keynote	The next major opportunity for first-party confirmation	Availability for every account, region or API customer

Until the keynote or a first-party release note says otherwise, the right answer to “when is OpenAI o released?” is expected around DevDay, not officially released.

What the long-horizon cloud agent is rumored to do

The consistent idea in the pre-event reports is long-horizon execution. Instead of answering once, o could keep a task alive in the cloud, use tools repeatedly, and continue after a user closes the app. Possible use cases include research, planning, software work, coordination and other multi-step personal tasks.

Those are product hypotheses, not supported capabilities. A real long-running agent must answer questions that ordinary chat marketing often skips:

• Can the user pause, cancel, resume and inspect every tool action?
• What happens when a task needs a login, payment, private file or human approval?
• Is execution measured in minutes, hours, days or weeks, and what are the limits?
• Does the agent keep running when the phone or browser is offline?
• Which integrations are first-party, and which require user-granted OAuth access?
• How are data retention, prompt injection, delegated authority and irreversible actions handled?

The product will be meaningful only if these controls are clear, not merely because it can keep a process alive in the background.

Aeon, GPT-6 Astra, and the “o” model question

One recurring claim says that Aeon is an Astra-derived variant that powers o. There is no official OpenAI model card establishing that relationship. Even if an internal model is used, a product may route different subtasks to several models, tools or services; a reported internal codename would not tell developers what to put in an API request.

The name also creates confusion with OpenAI's older o1 and o3 reasoning models. A shared letter is not a technical lineage. o1 and o3 are model identifiers used for API or ChatGPT model families; the rumored o is described as a user-facing agent product. Wait for an official model ID and documentation before treating them as one stack.

OpenAI o price: what has and has not been reported

There is no confirmed price for o. The reported $500/month ChatGPT Pro Max plan is a separate, unconfirmed rumor, described as a possible low-latency subscription that may accompany the agent announcement. It is not an API rate, and it does not establish that o will cost $500 per month.

These are the OpenAI API routes currently available through APIMaster, not prices for o:

Current API option	Original input / output	APIMaster starting input / output	Active routes	Use it for
gpt-6-sol	$2 / $10 per 1M	$0.089307 / $0.446535	11	Complex coding and agent workflow steps
gpt-6-astra	$10 / $50 per 1M	$0.446535 / $2.232675	15	Hardest reasoning and escalation tasks

The prices were checked September 28, 2026 and can change with route supply. They are not a forecast of an o subscription, hosted-agent fee or future OpenAI API price. Use the Sol card or Astra card before sending production traffic.

Is the donut-shaped hardware real?

Another part of the rumor describes a consumer device shaped like a donut, with moving or active parts, intended to work alongside o. This has not been announced by OpenAI. A product image, industrial-design description or code trace would not establish a shipping device, a release date, a price or compatibility with the agent.

If hardware appears at DevDay, verify whether it is a prototype, developer kit, research demonstration or retail product. The important details will be microphones and cameras, local versus cloud processing, account binding, physical mute controls, battery life, data retention and what happens when the network is unavailable.

How o would differ from a normal API agent

An API agent is an application that your team owns: you define the system prompt, tools, memory, approvals, storage, retry policy and budget. A personal resident agent would move more of that orchestration into a provider-managed product.

Question	Your API agent	Rumored OpenAI o
Orchestration	Your application controls the loop	OpenAI product would control more of the loop
Persistence	Your database and job queue	Reportedly provider-hosted background execution
Approval policy	You implement it	Must be explained by OpenAI before trust
Model selection	Explicit API model ID	Not published
Cost	Token, tool and infrastructure usage	Not published; $500 plan is only a rumor
Portability	Usually replaceable behind your interface	Could be tied to ChatGPT account and integrations

Teams should not wait for a rumored consumer product to build basic long-running workflows. Keep tasks idempotent, record tool calls, require approval for external side effects and make the model configurable. Those practices will help whether o arrives or not.

What developers should verify at DevDay

When the keynote happens, look for first-party answers to these questions rather than relying on reposted screenshots:

1. Is the product officially called o, and is it a model, an agent, a plan or a device ecosystem?
2. Is there a public release date, waitlist, region restriction and account tier?
3. Is there an API model ID, or is o available only inside ChatGPT?
4. What are the task-duration, context, output, tool and concurrency limits?
5. Can users inspect, pause, cancel, approve and roll back actions?
6. What integrations, browser controls, file access and OAuth permissions exist?
7. What is the real price, and is any $500 plan separate from API billing?
8. Is Aeon an official model name, an internal codename or unrelated speculation?

Do not update a production integration from the first social post. Wait for the model reference, pricing page, safety documentation and a successful low-risk test.

How to use an OpenAI model through APIMaster today

If the goal is coding or a multi-step API workflow today, create an APIMaster account, create an API key, set the OpenAI-compatible base URL to https://apimaster.ai/v1, and start with gpt-6-sol:

curl "https://apimaster.ai/v1/chat/completions" \
  -H "Authorization: Bearer $APIMASTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-6-sol","messages":[{"role":"user","content":"Inspect this failing test and propose a minimal patch."}]}'

This is a currently callable API model, not the rumored o product. For higher-stakes escalation, switch the model value to gpt-6-astra after checking its live card and route compatibility. Keep the model ID configurable so an official o API, if announced, can be tested without rewriting the worker.

FAQ

Is OpenAI o officially released?

No. As of September 28, 2026, o is a pre-DevDay product rumor. OpenAI has posted a DevDay countdown, but has not published an official o product page, API model ID, price or availability notice.

Is OpenAI o the same as Meta Muse?

No. The reports use Meta Muse as a comparison point for a persistent agent concept. o is described as an OpenAI product rumor; it is not Meta Muse and there is no evidence that the two are the same system.

Is OpenAI o the next o1 or o3 model?

That is unconfirmed. The letter overlap does not prove a shared model family. The reports describe o as a user-facing agent product, not an announced API model that replaces o1 or o3.

Will OpenAI o run for days or weeks?

That is one of the central claims in the reports, but no official task-duration limit has been published. Wait for product documentation to confirm whether long-running work is actually supported and under what controls.

Is ChatGPT Pro Max really $500 per month?

The reported $500/month ChatGPT Pro Max plan is unconfirmed. It should not be treated as an official price for o, an API rate or a guarantee that the plan will launch.

Is Aeon the model behind OpenAI o?

No official source confirms that. Aeon is described in reports as an Astra-related variant or internal name; until OpenAI publishes a model reference, developers should not use aeon as an API model ID.

Can I call OpenAI o through an API today?

No public o API route was available in the September 28 check. You can use the current gpt-6-sol or gpt-6-astra API options through APIMaster while waiting for first-party details.

Sources

• OpenAI DevDay — official event page and the first-party event context to check after the keynote.
• OpenAI Developers on X — the account that posted the September 27 DevDay countdown referenced in the pre-event reports.
• GPT-6 Astra model reference — current documented OpenAI API baseline; it does not establish an o product.
• OpenAI API pricing — official API pricing reference for released models.
• APIMaster live GPT-6 Sol card and GPT-6 Astra card — current route prices checked September 28, 2026.

For a production-ready API today, register with APIMaster, create a key, and test gpt-6-sol through https://apimaster.ai/v1. Return to this page after DevDay for any first-party o release details.