Note: All market prices are as of 19 August 2026. Funding figures are as reported at the time of each round. This is a fast-moving file and several of the sub-markets below are very thinly traded, which is flagged where it matters.
1. Two coin flips, two and a half times apart
The cleanest way to compare two valuation ladders is to find where each one crosses 50 percent.
Anthropic's crossing point is $2.5T, priced at 49.5 percent for 31 December. OpenAI's is $1.0T, priced at 60.5 percent. That is a 2.5x gap between two companies that were within $113B of each other in late March.
Neither number is a valuation. Each is the market's median outcome for where a private mark lands by year end, and the distance between them is the story.
2. The $965B round is already behind the market
Anthropic raised $65B in a Series H led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia, closing at $965B post-money on 28 May. That round nearly tripled the $380B valuation set in February, which itself was more than five times the $61.5B mark from March 2025.
The board treats that as history. Trading above $1.1T is settled at 100 percent. Above $1.5T prices at 92.5 percent. Above $2.0T at 66.5 percent.
The downside legs are worth reading too: back below $800B prices at 8.5 percent, and below $600B at 4.9 percent. The market is not pricing this as a fragile mark.
3. An OpenAI mark-down is a live question
OpenAI raised $122B in late March at a $852B post-money valuation, the largest private round on record at the time. The market prices above $900B at 80 percent and above $1.0T at 60.5 percent.
Then the other side. Back below $800B prices at 32.5 percent. Below $750B at 24.5 percent. Below $700B at 18.5 percent.
Roughly a one in three chance that the most valuable AI company of March is worth less by December than it was then. Anthropic's equivalent contract prices at 8.5 percent. That asymmetry, rather than either headline valuation, is the sharpest thing on this board.
4. Confidence decays as the horizon lengthens
Anthropic holds the best-model contract at 97.9 percent for the end of August, 90.5 percent for the end of September and 66.5 percent for the end of 2026.
That decay is not fading conviction about the present. It is the market pricing how much can change in four months in a field where frontier releases land every few weeks. A 97.9 percent for a two-week horizon and a 66.5 percent for a four-month one are consistent readings of the same position.
For anyone building a view, the useful number is the one matching your own time horizon. Quoting the two-week figure as evidence about next year overstates it, and quoting the year-end figure as a read on today understates it.
5. The runner-up is not the one you would guess
On the end-of-2026 book, Anthropic prices at 66.5 percent, xAI at 13.1 percent, OpenAI at 9.0 percent, Google at 8.0 percent and Meta at 3.1 percent.
OpenAI sits third on a market about its own core product, four points behind xAI and one point ahead of Google. That is a striking placement for the company that defined the category, and it is the single most contrarian number in this article relative to mainstream coverage.
Caveat worth stating: this book carries $721K in total, with roughly $59K on the xAI leg and $59K on the OpenAI leg. It is real money but not deep money, and the ordering between second and fourth place should be treated as soft.
6. GPT-6 is priced as a fourth-quarter event
GPT-6 by 21 August prices at 1.1 percent. By 31 August, 1.9 percent. By 15 September, 13.5 percent. By 31 October, 57.5 percent. By 31 December, 80.5 percent.
The shape says the market expects the release and expects it late. Nothing meaningful before mid-September, a jump through October, and four in five by year end.
That timing matters for point 5. If GPT-6 lands in October or November, the end-of-2026 leaderboard has time to reprice before it resolves, which is part of why that book sits at 66.5 rather than 97.9.
7. The listing race is priced tighter than the valuations
An OpenAI IPO by 31 December 2026 prices at 20.5 percent. Conditional on it happening, a closing market cap above $1T prices at 76 percent and above $1.4T at 43.5 percent.
Anthropic filed confidentially on 1 June. Its bracket market prices no listing by the end of 2027 at just 3.2 percent, with the modal outcome a closing capitalisation between $2.0T and $2.25T. A separate lower-bracket book puts above $600B at 95 percent.
Note the structure, because it repeats a lesson from the token launch markets: the closing-cap contracts are conditional on a listing happening. Reading OpenAI's 76 percent as a straight forecast of a $1T listing ignores the 20.5 percent gate in front of it.
8. The DeepSeek narrative is finished
Best Chinese AI company at the end of August prices Alibaba at 97.2 percent, Moonshot at 1.6, Z.ai at 0.7 and DeepSeek at 0.4.
Eighteen months ago DeepSeek was the story that moved US equity markets. It now prices below one percent on its home board, behind two companies most Western coverage has never mentioned.
This is a useful check on narrative persistence. The DeepSeek moment produced an enormous volume of analysis and a lasting mental model, and the market that actually resolves on it has moved on completely.
9. Key-man risk prices as a tail everywhere
Sam Altman being out before 2027 prices at 7.5 percent on $94.9K. Sundar Pichai prices at 8.5 percent and Andy Jassy at 8.2 percent on the same book.
After a year in which OpenAI leadership speculation has been a recurring news cycle, the market prices Altman as marginally less likely to depart than the CEOs of Google and Amazon. Attention and probability are different currencies, which is the same lesson the alien market teaches at a much larger scale.
10. Read the volume before the level
Some of the most quotable numbers here sit on almost nothing.
The Anthropic leg of the end-of-August book carries $441.7K and the live OpenAI IPO leg carries $909.2K. Those are real. But the GPT-6 contract for 31 October, priced at 57.5 percent, carries $775. The 30 November leg at 63 percent has never traded at all.
So the GPT-6 ladder in point 6 is directionally useful and numerically unfunded past September. Use it to establish that the market expects a late release. Do not quote 57.5 percent as though it has been tested.
The same discipline applies across this board: the valuation ladders and the near-dated leaderboards are deep, and almost everything with a distant date is thin.
What this adds up to
The market has made a specific and unfashionable call. It prices Anthropic ahead on model quality at every horizon, prices a further doubling of its valuation as a coin flip, and prices a one in three chance OpenAI is marked down before year end. It puts xAI ahead of OpenAI for the end of 2026, and it puts DeepSeek below one percent in China.
None of that is a prediction. It is where the cost of disagreeing currently sits, and on several of these books that cost is set by a fairly small number of participants.
Two things worth watching. Whether the GPT-6 release lands inside the October window the market expects, because the end-of-2026 leaderboard has been left room to reprice around it. And whether OpenAI's downside legs move, since a mark-down at 32.5 percent is the most consequential number here and the one most likely to be resolved by a single announcement.
Where to watch it
Every market in this article is live on ApeX Omni, where Polymarket's order books are integrated directly: the same liquidity and the same settlement source, minus the KYC. The board shows the full valuation ladders, the leaderboard at each horizon and the volume behind every leg, which is what makes it possible to tell a tested price from a quoted one.
We appreciate your continued support.
Published by ApeX. For informational and educational purposes only. Not investment advice, and not a forecast or endorsement of any outcome. Trading involves risk, including total loss.
