Release-timing contracts
Whether a named model or product ships inside a window. Preview versus general availability is the usual dispute.
AI · Live markets
AI benchmarks, safety and tech market signals.
LLM agents, frontier safety, AI infrastructure and automation. Use this board to follow sourced stories, compare platform fit and check resolution quality before opening any market.
Related boards: Esports · Weather · World Cup · Industry events · Event betting · Finder · Fees and risk
Live AI markets
Updating from Polymarket.
Active AI markets with volume, prices and related platforms. Auto-refreshes every 5 minutes.
Loading live markets…
Settlement
How AI markets resolve.
Read the resolution source before the price — it decides more outcomes than direction does.
AI markets are the hardest category on this site to price, because almost nothing resolves against a number that already exists. A football match produces a score; a model release produces a blog post, a system card and an argument about whether the thing that shipped is the thing the market described.
That makes wording the whole trade. Two markets asking “will a frontier lab release a new flagship model this quarter” can settle differently depending on whether a preview, a staged rollout or an API-only release counts. Read the resolution paragraph before the probability.
| Settlement source | What to check |
|---|---|
| Lab announcementsModel launches, version numbers, deprecations | Official blog or system card, not a press summary or a leak |
| Public leaderboardsBenchmark thresholds and ranking claims | Boards are versioned and re-scored; check which snapshot the rules name |
| Regulatory textsEU AI Act phase-ins, national AI rules | Statutory dates are firm; enforcement guidance often slips |
| Company filingsCapex, data-centre and compute commitments | Quarterly cadence, so long-horizon contracts settle late |
Contract types
What trades on this board.
The same headline event can be listed as several different contracts. These are the shapes you will meet in ai.
Benchmark-threshold contracts
Whether a score crosses a stated line. Self-reported numbers and independent evaluations frequently disagree.
Regulation contracts
Whether a rule takes effect, is delayed or is amended by a stated date.
Infrastructure contracts
Compute, capex and data-centre milestones, usually settled from filings or official statements.
Calendar and liquidity
- No fixed calendar. Liquidity follows launches, not a season.
- Research conferences and major developer events cluster announcements.
- Quarterly earnings settle most compute and capex questions.
- Regulatory phase-in dates are the only genuinely predictable catalysts.
Where ai contracts go wrong
- “Frontier”, “flagship” and “AGI” are marketing terms, not resolution criteria — if the rules do not define them, skip the market.
- Benchmarks get re-run and re-scored; a contract naming a live leaderboard can move without any model changing.
- Staged and region-limited rollouts break simple “released” wording.
- Long-dated AI contracts tie up capital for quarters with wide spreads.
Venue fit
Where ai contracts actually list.
Coverage differs sharply by category. These are the venues worth comparing first here — always verify eligibility yourself.
MetaculusThe deepest AI question set, with written rationales. Reputation-based rather than cash, so treat it as research.
ManifoldWidest AI coverage and free to use, which makes it the cheapest way to learn how these questions get worded.
PolymarketWhere headline AI questions get real money and tighter spreads, subject to regional eligibility.Editorial comparison only, not financial advice. Prediction markets can lose money, regional eligibility changes, and settlement rules are set by each venue. Read our rating method and fees and risk guide before funding an account.
FAQ
AI settlement questions.
How do AI prediction markets actually resolve?
Almost always from an official lab publication, a named public leaderboard snapshot or a statutory text. Press coverage and leaks do not settle a contract, even when they turn out to be correct.
Why do two AI markets on the same launch disagree?
Because they define the launch differently. One may accept a limited preview while another requires general availability, so the same event settles Yes on one venue and No on another.
Are AI markets a good starting category?
No. Resolution wording is looser than in sports or macro, and horizons are long. Learn contract mechanics on a dated, single-source category such as economy or weather first.
Need a faster answer?
Use the swipe chooser or compare tool to shortlist platforms by country, KYC, crypto support and event category.