Manifold traders give “AGI before 2027” just a 2% chance, even as OpenAI declares the “AGI era.” A play-money market on Manifold asking “Will we get AGI before 2027?” is sitting at roughly a 2% implied probability of YES as of September 17, 2026, with 333 unique bettors and one of the heaviest 24-hour volumes on the platform. The activity spike lines up with two real events this month: OpenAI’s launch of GPT-6 Astra on September 3, and Anthropic chief executive Dario Amodei’s September 12 essay arguing the industry should deliberately slow down. The gap between corporate messaging and crowd money is the story.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What this market is actually asking
The market, hosted on Manifold, poses a deceptively simple binary: will humanity reach artificial general intelligence before January 1, 2027? Manifold runs on play money rather than cash, so it is not a regulated exchange like Kalshi or an offshore venue like Polymarket. What it offers instead is a fast, liquid read on what a self-selecting crowd of AI-watchers, researchers, and hobbyist forecasters actually believes when they have to put a number on it.
Because “AGI” has no single agreed definition, a market like this leans heavily on its own resolution criteria and on the judgment of whoever settles it. That ambiguity is not a bug here. It is precisely why the question is interesting: traders are not just forecasting technology, they are forecasting whether any credible authority will declare the milestone reached in the next roughly three and a half months. If you are new to how these contracts translate belief into a price, our primer on how prediction markets work lays out the mechanics.
The odds, and what 2% implies
As of September 17, 2026, the YES side trades near a 2% implied probability on Manifold. In plain terms, the crowd is pricing roughly 49-to-1 against AGI arriving before the new year. That is close to the floor for a market that still has meaningful volume, and it sits well below the timelines that the AI labs themselves publish.
An implied probability is simply the market price read as a percentage chance. A contract trading at 2% means the marginal buyer thinks there is about a 1-in-50 shot the event resolves YES. For a fuller walkthrough, see our explainer on implied probability. The notable feature is not the low number by itself, but the volume behind it: heavy 24-hour trading and 333 distinct participants means the 2% is a considered consensus, not a stale quote nobody is defending.
What moved it this month: GPT-6 Astra
On September 3, 2026, OpenAI unveiled GPT-6 Astra, its most capable model to date, and framed it in unusually strong terms. According to CNBC and Axios, the company said the system may represent the early stages of artificial general intelligence. OpenAI president and co-founder Greg Brockman closed the launch session with a line that ricocheted across the industry: “Welcome to the AGI era.”
The model itself is built around what the industry calls computer use. As Fortune reported, Astra is designed to operate software the way a person does, working across browsers, spreadsheets, websites, and desktop applications, and carrying out multistep workflows rather than merely describing them. OpenAI staged the rollout from a limited set of organizations to ChatGPT Plus, Pro, Business, and Enterprise tiers, plus the API, with input priced at $10 per million tokens and output at $50 per million. OpenAI also pointed to a specific number to back the framing, reporting that Astra reached 99.9% on the ARC-AGI-3 reasoning benchmark in September 2026, and that figure, more than any single prior release, is what put “AGI” back at the top of the news cycle and drew fresh bettors to the Manifold contract.
Amodei’s “pace the frontier” essay
The second catalyst pulled in the opposite direction. On September 12, 2026, Anthropic CEO Dario Amodei published an essay titled “We Must Pace the Frontier,” arguing that the industry should slow the rate at which it improves raw model capability so that alignment and safety work can catch up. In his own post announcing the essay, Amodei said Anthropic would unilaterally commit to the first step of a three-part plan: giving third-party evaluators permanent, employee-level access to its systems.
What made the moment land was the response. Within hours, OpenAI’s Sam Altman publicly agreed with the general premise that the frontier should be paced. Two companies that compete directly for the same enterprise customers and the same capability headlines aligning, at least rhetorically, on a slowdown argument is rare. The irony sharpened less than three weeks later: on September 30, 2026, Google launched its own flagship model, Gemini 4 Argon, pricing API access at $2 per million input tokens and $10 per million output tokens, undercutting OpenAI’s rates even as the industry’s pacing debate was still playing out. For a market betting on an imminent AGI declaration, a coordinated “let us slow down” message from the two loudest labs reads as a reason to fade the hype, not chase it, though a third major lab racing to ship a cheaper, competitive model suggests the slowdown talk may be more rhetorical than universal. We covered the broader safety turn in our report on the OpenAI, Anthropic, and Google safety talks.
The definition problem doing the heavy lifting
The core reason the market can price 2% while a launch stage host says “welcome to the AGI era” is that the two statements use “AGI” to mean different things. OpenAI’s own working definition describes highly autonomous systems that outperform humans at most economically valuable work. A model that can competently drive a browser and fill a spreadsheet is a long way from clearing that bar across most of the economy.
According to a Bloomberg feature published September 4, Anthropic is the only major AI company that publishes official AGI timelines, and it expects the milestone by early 2027. The company’s own release pace backs up that stated urgency: on September 22, 2026, Anthropic shipped Claude Opus 5.5, pricing API access at $4 per million input tokens and $20 per million output tokens. Altman, for his part, has said OpenAI is “not quite yet” at AGI but expects an internal system he would personally call AGI by the end of 2026. Note that even the most aggressive lab timeline lands in early 2027, which is on the far side of this market’s January 1, 2027 cutoff. That single detail explains much of the distance between the marketing and the money.
Who and what is involved
The named actors here are concrete. On the accelerationist side of the messaging sit Greg Brockman and Sam Altman of OpenAI, whose GPT-6 Astra launch supplied the “AGI era” language. On the caution side sits Dario Amodei of Anthropic, whose essay reframed the month around pacing and third-party evaluation. Bloomberg’s reporting places both firms in a direct race, with Anthropic uniquely willing to attach a public date to the goal.
The market itself was created by a Manifold user and settles according to that market’s stated criteria, which is a reminder that play-money venues aggregate opinion rather than deliver an official verdict. The 333 bettors are the crowd doing the pricing. None of them control whether a lab issues a formal AGI declaration, which is part of why the number stays low: an event that depends on a specific, contested announcement inside a narrow window is inherently hard to bet YES on.
How this market compares
The Manifold contract is not the only place traders are pricing AGI. Estimates vary widely by venue, timeline, and exactly what is being asked, which is itself the lesson.
| Venue / source | Question framing | Timeline | Implied estimate |
|---|---|---|---|
| Manifold (this market) | Will we get AGI? | Before 2027 | ~2% YES (Sept 17, 2026) |
| Polymarket (reported) | OpenAI achieves AGI | By 2027 | ~9% (Aug 2026) |
| Kalshi (reported) | OpenAI achieves AGI | By 2030 | ~45% (Aug 2026) |
| Anthropic (official) | Company AGI timeline | Early 2027 | Stated expectation |
| OpenAI (Altman) | Internal system he would call AGI | End of 2026 | Stated expectation |
The pattern is consistent: the tighter the deadline, the lower the market-implied odds, and the further out the horizon, the more the crowd is willing to entertain it. For a sense of how these venues differ in structure and regulation, see our Kalshi vs Polymarket comparison.
What to watch next
Three things could move this contract before it resolves. First, any formal declaration from OpenAI, since Altman has flagged an internal system he might call AGI by year-end, and a public version of that claim would test the market’s resolution criteria directly. Second, follow-through on Amodei’s pacing proposal: if other labs commit to third-party evaluator access, the “slow down” narrative strengthens and YES odds likely stay pinned. Third, benchmark and independent-evaluation results for GPT-6 Astra, which will either validate or deflate the “AGI era” framing over the coming weeks. Fourth, how those Astra numbers hold up against Google’s competing claims: Gemini 4 Argon posted a 77.9% score on the DeepSWE v1.1 coding benchmark in September 2026, giving traders a concrete rival data point to weigh against OpenAI’s self-reported figures.
For readers tracking how event-driven prediction markets swing on headlines more broadly, our coverage of the Fed rate-decision odds on Polymarket shows the same dynamic in a very different domain.
Frequently asked questions
Why is the market only at 2% if OpenAI says we are in the “AGI era”?
Because the two are measuring different things. The “AGI era” line is framing around a new model’s capabilities, while the market asks whether a full, economy-wide AGI milestone will be declared before January 1, 2027. Even Anthropic’s public timeline lands in early 2027, just past the cutoff.
Is Manifold real-money betting?
No. Manifold uses play money, so it is not a regulated financial exchange. It aggregates the opinions of an engaged forecasting crowd, which can be informative, but it does not carry cash stakes the way Kalshi or Polymarket do.
What is GPT-6 Astra?
It is OpenAI’s most capable model as of its September 3, 2026 launch, built around computer use: operating browsers, spreadsheets, and desktop apps to complete multistep tasks rather than only describing them.
What did Dario Amodei actually propose?
In his September 12 essay, he argued for deliberately pacing capability gains so safety work can keep up, and committed Anthropic to giving third-party evaluators permanent, employee-level access to its systems.
The Bottom Line
Sources
- CNBC — OpenAI announces rollout of GPT-6 Astra model
- Axios — OpenAI releases GPT-6 Astra, says it may represent AGI
- Fortune — OpenAI debuts GPT-6 Astra, Brockman says start of AGI
- Bloomberg — What Is AGI? OpenAI, Anthropic Race for Artificial General Intelligence
- Dario Amodei — “We Must Pace the Frontier” announcement
- VentureBeat — ‘Welcome to the AGI era’: OpenAI launches GPT-6 Astra
- Manifold — Will we get AGI before 2027?
Prediction markets carry risk and are not investment or betting advice. Availability is restricted by jurisdiction: Polymarket is not available to US persons, and Kalshi is a CFTC-regulated US exchange. Manifold uses play money. Participation is limited to those 18+ or 21+ as applicable. If gambling is a problem for you or someone you know, call 1-800-GAMBLER for confidential help.
Related Coverage
- Superconductor Bet Heats Up: Manifold Traders Give a Room-Temperature Breakthrough 15% Odds Before 2031
- Trump on US Currency? Manifold Traders Price 97% Odds His Face Appears Before 2028
- Will Ghent Fire Nathan Cofnas Before 2028? Traders Pile Into Manifold Bet at 22%
- Roon's 'Great Minds Discuss Flops' Tweet Is a Manifold Market, and Traders Bet 9% It Ages Well
- Manifold Traders Pile Into 'Fed Rate Hike in 2026?' as Odds Hit 94%
