By Ved Vyas | Published 29 Sep 2026

Level up your AI workflow with premium prompt packs, guides, and creative AI resources. 🚀
Explore the full collection and grab your favorites here: AI Unfiltered on Gumroad

OpenAI gave its new agent a cartoon face and a name that's awkward to search for. The face isn't the story. On the day Dots launched, OpenAI halved two allowances on the $200 Pro plan, added a $500 plan above it, and made the first Dot free for anyone eligible.

OpenAI Dots are always-on personal AI agents that launched inside ChatGPT on 29 September 2026. They run on GPT-6 Astra, pursue goals you set while you're away, and reach you through Slack, Microsoft Teams, and soon SMS. Pro and Business Premium users get one named Dot, and chatting with it reportedly doesn't consume ChatGPT usage limits.

One row in that table changes what a Dot costs you in practice, and it sits beside a plan change most launch stories buried. I'll get there after the basics.

A chatbot waits. OpenAI Dots don't. That's the whole difference, and the rest of this section is detail. OpenAI's pitch, as Engadget reports it, is work done your way sometimes before you even think to ask.

You assign a task in plain language, attach files or photos if you need to, and the Dot keeps going between conversations. NewsBytes describes recurring schedules, proactive updates, and memories built from the apps you connect, and you choose how much rope the agent gets, from asking before every action to acting entirely on its own.

TechCrunch lists two example jobs for OpenAI Dots from the launch material: a developer has a Dot watch customer feedback and ship bug fixes, and a scientist has one rerun analyses and chase down odd experimental results. Background work with a clear finish line. That's where agents earn their keep.

TechCrunch also calls the system hardware-agnostic, meaning the same agent follows you across surfaces instead of living inside one app. The longer plan is teams of Dots, each provisioned with its own identity, credentials, and tools, so one person could eventually run a small crew of specialists that report back to a single primary agent.

You name your primary Dot, and the branding is what TechCrunch called "little floating cartoons." That styling matters. It's the reason so much of the online reaction is about a mascot instead of a model, and the Hacker News thread later in this piece shows it.

Nobody outside OpenAI has said what OpenAI Dots can do at scale. Everything here is description. Nothing is measured.

Pro and Business Premium subscribers get OpenAI Dots first, and Enterprise access is reported by only two outlets. Regional limits apply. Europeans and Brits will feel that one. NewsBytes says Pro users outside the EEA, Switzerland, and the UK can turn one on, while Business Premium works in every supported region.

The outlets disagree on the details. I lined up what each one says.

OfficeChai's line conflicts with everyone else's, and since its article also covers ChatGPT Space, it may be describing that product's rules instead of the rules for Dots. I can't tell from the article. My working reading is that Pro and Business Premium are solid, Enterprise is likely, and the regional carve-out matters if you're in Europe or the UK.

Check the ChatGPT pricing or help page before you plan around any of this, because OpenAI's own page is the tiebreaker and I couldn't load it.

The OpenAI Dots price is zero on top of an eligible plan, according to Engadget, and conversations with a Dot don't count against ChatGPT usage limits on either Pro tier, and OfficeChai reports the same exemption in different words.

Here's the part most stories skipped. The same day, OpenAI changed the plans around it.

Existing $200 subscribers keep their old limits for a while and get a one-time credit to offset the new ones, per Engadget, though the report doesn't say how long a while is. The $500 tier's name is reported inconsistently: Engadget says "Pro 500," pre-launch reports said "Pro Max," and one outlet mentioned a rename to "Pro Standard and Pro More."

Now the catch. It reads as covering conversations. Engadget's launch story also says standard usage limits still apply in Codex and Work apps, where Dots can also be launched from. Nothing I read says whether a Dot's background work inside Codex draws down that halved allowance, or whether it gets the same exemption as chat. If it counts, free gets narrow fast.

OpenAI Dots can work across a very large set of apps, run tasks in the cloud, and use a browser or a PC. Engadget cites access to more than 4,000 apps through OpenAI's plugins, with context carried across desktop, web, and mobile.

Safety is where the OpenAI Dots launch material spends real effort. Passwords get handled without ever exposing credentials to the underlying model, proactive research runs under read-only safeguards, and approval rules are customizable, with built-in restrictions layered on top of whatever you set. Enterprise buyers get Microsoft's Agent 365 controls, per TechCrunch and Whalesbook.

OfficeChai adds that a Dot can take calls, email, and voice, and can make purchases on your behalf. Only that one outlet says so. Purchasing stays unconfirmed until OpenAI's page or a hands-on report backs it up, and I wouldn't hand any agent a card before that happens.

Reach is broad on paper. Slack and Teams work now. SMS is next.

The model underneath matters for cost. OfficeChai published a pricing table alongside the launch, and it lists Astra at $10 input, $1 cached input, and $50 output, against $2, $0.10, and $10 for the cheaper Sol model. By my arithmetic that's 5x on input, 10x on cached input, and 5x on output, though the summary I read doesn't state the unit and API prices are normally quoted per million tokens. Either way, every OpenAI Dots task runs on the pricier of the two models, which is a reasonable explanation for why OpenAI would want Dot usage tracked separately from ordinary chat.

Not yet. None that I could find. The launch happened around 10 am Pacific on 29 September, and every source I read describes announced features, not tested ones. This page isn't a review either. Real numbers go here when they exist.

Meta's Muse got there first. Semafor reports more than 3 million downloads for Muse and frames Dots as OpenAI's answer, with the same cute, nameable-mascot approach. Sam Altman tried Muse and called it "a nice product," per Semafor, and voiced no worry about the competition.

Both products share a failure pattern. Semafor says Amazon blocked Muse from shopping on its platform, and that Meta's agent gave out a Toronto resident's home address on Facebook Marketplace without consent. A Dot inherits those risks the moment it can act for you, and no amount of cartoon charm changes that.

The security designs differ on paper. Wccftech says Muse runs inside isolated virtual machines with dedicated browsers to protect your data and credentials. OpenAI's answer, per Engadget, is to keep passwords away from the underlying model and to put proactive research behind read-only safeguards. Neither design has been tested in public.

Wccftech, citing a JPMorgan benchmark, says Muse scored 9.3 and answered in about 7 seconds against roughly 38 for rivals. I couldn't confirm the benchmark. Hold it loosely. Nobody has benchmarked a Dot against Muse, so that 9.3 tells you nothing about how a Dot performs.

One more detail: 9to5Mac's pre-event teaser framed the launch as ChatGPT's reply to agents from Meta and SpaceX, naming Muse and something called Grok Bot. So OpenAI Dots arrive in a race.

No fair comparison exists yet. What you have is two launch announcements, one head start, and a very large amount of marketing energy from both companies.

Mostly skepticism. The Hacker News thread on OpenAI Dots leans negative. The complaints repeat.

Several commenters compared the cartoon styling to Microsoft's old Office paperclip and called it a strange choice for something with access to your accounts. One said the lines between Codex, ChatGPT Work, and Dots are getting blurry, which is fair, since TechCrunch says you can launch a Dot from either Codex or ChatGPT. Others said the announcement page never made clear what the product is, which is a rough thing to hear about a launch page. The missing EU and EEA rollout got read as a compliance warning, and people raised the risk of an agent with wide system access.

Cost came up too. Commenters read Dots as a Pro-plan feature that's expensive for ordinary consumers, with an unclear pricing model, and a few said the demo felt slow. That last complaint is worth watching, because an always-on agent that's slow at its own launch demo has a hard time selling "before you even think to ask."

Skepticism at launch isn't a verdict. Early threads skew negative on nearly everything. Still, "what is this, exactly?" from technical readers is a signal, and it's the same confusion that sends people searching for an OpenAI Dots explainer today.

The mascot will get the jokes. The pricing will get the consequences, and those will land on your invoice a lot sooner than any agent will finish a project for you.

On one day OpenAI told Pro users a Dot is included and doesn't eat ChatGPT limits, while halving the Codex and Work allowance and the weekly GPT-6 Pro cap on the $200 plan. A new $500 tier now sells the top allowance plus a mode that runs up to 300 tokens per second in Codex and Work apps. My read is that the Dot is the reason to stay on Pro, and the cuts are the reason to consider paying more.

The strongest counterpoint is fair. If you mainly want an agent doing background chores, a $200 plan with a free Dot may be a better deal than the plan was last week. That's a real perk.

I still land on the other side. Anyone who spent the $200 plan on heavy Codex days just lost half their allowance, and the Dot exemption doesn't give it back. Nothing says the exemption is permanent. Until OpenAI states how Dot work inside Codex is counted, I'd budget as if the exemption might not cover it and treat any surplus as a bonus. That's a cautious call, and if OpenAI's page settles it in the generous direction, I'll happily revise this section. The OpenAI Dots price question deserves a straight answer from OpenAI itself instead of a stack of press summaries, mine included.

Early reports had the name wrong. Before launch, Wccftech called the agent "o," put DevDay on 26 September, and tied it to a rumored donut-shaped hardware device. The event ran on 29 September. The product is called Dots. NewsBytes writes it as singular "dot" in its headline, while TechCrunch, Engadget, and Semafor use "Dots." If you're searching, try both spellings.

Pre-launch tier names were off too. Reports called the $500 plan "Pro Max," and Engadget's launch-day story calls it "Pro 500." Copy the name from OpenAI's pricing page, not from a rumor post.

Here's what I couldn't confirm:

Bloomberg's headline mentions the $500 tier, but the article was blocked to my tools, and CNBC's live blog returned an error. I'll update when I can read either. Two other pages I pulled added nothing: one syndicated only a headline and the other never mentions Dots, a reminder that a lot of launch-day coverage is thin.

A Dot is included with an eligible plan, per Engadget, so there's no separate fee, but the plan itself isn't free: Pro is $200 a month and a new $500 tier sits above it. Chats with a Dot reportedly don't count against ChatGPT limits, but standard limits still apply in Codex and Work apps, where Dots can also be launched from.

Pro and Business Premium include OpenAI Dots in nearly every report. Engadget and OfficeChai also mention Enterprise, though OfficeChai's wording conflicts with the others. Every Pro subscriber gets one Dot, on both the $200 and $500 tiers. Confirm on OpenAI's pricing page.

Not for Pro users, according to NewsBytes. Pro access is limited to users outside the European Economic Area, Switzerland, and the UK. Business Premium has no carve-out. OpenAI hasn't said, in anything I read, when or whether Pro access will open up there, so European Pro subscribers shouldn't plan around it.

Not yet. Each user gets one customizable, named primary Dot at launch, per Engadget. OpenAI says it plans teams of Dots working together on your behalf, and TechCrunch says individual Dots can be given specific identities, credentials, and tools. No date yet.

You need the ChatGPT desktop app or the desktop web version, per NewsBytes, and TechCrunch says you can launch a Dot from Codex or ChatGPT. After that you name it, connect apps, and set approval rules for whether it asks first or acts alone.