What OpenAI’s ‘o’ Assistant Leak Actually Tells Us

A reported ChatGPT code string labels “o” an “always-on assistant,” while a translator note identifies lowercase “o” as the product name. Reports also say the name briefly appeared on a ChatGPT Pro upgrade screen. These are meaningful product clues, but they do not confirm a public launch or explain exactly what the assistant can do. 18

What the leak suggests about “o”

“Always-on” points toward an assistant designed to keep working beyond a single chat. Some reports describe a possible agent that could continue tasks after a user closes the chat, but the leaked wording does not establish which apps it could use, what actions it could take, or how much user approval those actions would require. 138

A configuration reference to “o” as a display name and a reported “-o” email suffix add to the picture of a product being prepared. They are not, by themselves, proof that users will receive a dedicated email address or that the assistant will be available outside a test. 214

DevDay debut: possible, not confirmed

Pre-event reports pointed to OpenAI’s Sept. 29 DevDay as a possible announcement window. But the reporting available before the event said OpenAI had not announced “o,” and whether it would be revealed or made available to users remained uncertain. 210

That distinction matters: references in code and on an upgrade screen can suggest development, but they do not confirm a launch date, rollout plan or developer API access. 29

The rumored model and subscription price

One report claims that “o” would use a long-task model called Aeon, described as a variant of GPT-6 Astra. That model connection is a report, not something established by the product strings or a published model announcement. 5

The clearest pricing clue is a reported appearance on a $100-per-month ChatGPT Pro upgrade screen. It does not establish “o”’s final price, whether that screen reflects its eventual availability, or which plans would include it. 48

A separate $500-per-month Pro Max plan has also been rumored, but that is not confirmation of “o”’s price or whether the assistant would be included. 35 The available reporting does not settle the commercial terms.

How “o” could differ from Meta Muse and xAI Grok Bot

The comparisons are tentative because “o”’s capabilities have not been confirmed. Meta describes Muse as a personal agent that can work across a person’s apps using its own browser-equipped virtual computer. 44 By contrast, reporting describes Grok Bot as a group of persistent AI teammates built for multi-app work. 48

If “o” works as the “always-on assistant” label implies, it may sit closer to the personal-assistant idea behind Muse than to Grok Bot’s team-of-agents model. But there is not enough verified information to compare their actual capabilities, safeguards or performance directly. 184448

What OpenAI’s GPT-7 research focus does—and doesn’t—tell us

OpenAI applied research leader Boris Power was reported as saying that 80–90% of the company’s research is aimed at GPT-7, GPT-8 and later generations. 38 That statement concerns the direction of research; it does not identify the model behind “o” or verify the rumored Aeon connection. 538

The practical takeaway is to separate the clues from the speculation: the “always-on assistant” wording and reported Pro-screen reference point to a possible product, while its launch, precise capabilities, model and subscription terms remain unconfirmed in the available reporting. 1810