Menu
Tech

Will Google's next Gemini Pro model debut above 1500 on LMArena?

Resolution: Updated:

In short

The market treats this as unlikely. No Google model carrying the Pro label has cleared a 1500 score on LMArena's Text Arena before, and the contract only pays out if the very next one does it on its first day. That could change quickly if Google ships a model before the 1 January 2027 deadline that posts a sharp, visible jump over its predecessor.

Editorial illustration for: Will Google's next Gemini Pro model debut above 1500 on LMArena?

How the contract works

A contract on this question settles at $1 if the next Google model labeled Pro debuts on LMArena's Text Arena Overall leaderboard with a score of at least 1500, measured at 12:00 PM ET the day after it first appears; it settles at nothing if that model scores below 1500, or if no qualifying Pro model appears before 1 January 2027. The price at any moment is simply what buyers and sellers currently think the chance of that outcome is โ€” a contract priced at 0.30, for instance, would mean the market sees it happening about three times in ten, not that anyone has guaranteed it. A position can typically be sold before the resolution date at whatever the prevailing price is then, rather than being held to settlement.
What the market thinks happens
$100
Yes7%

The event happens

Costs now
$0.07
If you put in $100
$1,429
No93%

The event does not happen

Costs now
$0.93
If you put in $100
$108

Probability

History starts collecting once the event is tracked

How the price has moved

The only pricing data available for this market is its current reading of 8% on Polymarket, built from $70,236 in total volume, with no second venue to compare against and no published move-by-move history. A single-venue market at this volume level is better read as an early, thin-liquidity snapshot than as a settled consensus; the absence of a reported trigger for any recent move means none should be assumed. What the level itself communicates clearly is that traders see a low probability, consistent with the view that no Gemini Pro release has previously opened at a score this high.

Analysis

Context

LMArena runs a crowdsourced leaderboard where users compare two anonymous chatbot responses and vote for the better one; those votes feed an Elo-style rating, shown with 'style control off' in this market's rules, meaning formatting and verbosity are not adjusted for. Google has shipped a steady run of Gemini Pro releases on this leaderboard, each new generation typically improving on the last by a modest margin rather than a sudden leap. The question asks about whichever model Google next labels 'Pro' โ€” for example gemini-3-pro or gemini-3.1-pro-preview โ€” explicitly excluding Flash and Flash-Lite variants, which are smaller, faster models not meant to compete at the top of the rankings. Settlement is tied to a specific moment: the score shown on the leaderboard at 12:00 PM ET the day after that model first appears, with the window closing 1 January 2027.
The consensus across venues sits at 8%, drawn entirely from Polymarket, where this contract has traded $70,236 in volume. That is a thin market by prediction-market standards, and it trades on a single venue rather than several โ€” there is no second price to compare it against, so there is no spread to read for disagreement between pools of traders. A low volume figure like this means the 8% reading reflects a relatively small number of participants rather than a broad consensus, and it can move more on a single sizeable trade than a deeper market would. Still, the direction is informative: pricing this near one-in-twelve rather than near a coin flip says the market does not expect Google's next Pro release to clear a bar none of its predecessors have reached on debut. Two separate things have to happen for this to resolve Yes. First, Google has to actually ship a model it labels 'Pro' โ€” not just Flash or an experimental variant โ€” and get it onto the LMArena leaderboard before 1 January 2027; if the release slips past that date, the contract resolves No regardless of how strong the model eventually is. Second, that model's score at the specified 12:00 PM ET check has to sit at or above 1500, a level that requires either a genuine capability leap over the current generation or a shift in how LMArena's voting pool rates responses. Leaderboard scores have generally climbed across generations as models improve and the voting pool recalibrates, but a jump of the size implied here, landing on the very first day of eligibility, is a higher bar than the gradual drift seen in past Gemini releases.

What moves the probability

  1. Release timing before the deadline

    Google must ship and get a Pro-labeled model onto the LMArena leaderboard before 1 January 2027 for this to have any chance of resolving Yes. Any slip past that date resolves the contract No automatically, independent of how the model eventually performs. This timing constraint alone explains much of the current low pricing.

  2. Size of the score jump required

    LMArena ratings tend to move in moderate increments between generations rather than sudden spikes. Reaching 1500 on debut implies either an unusually large capability gain or a recalibration of the rating pool as new, stronger competing models enter the comparisons, both of which are harder to predict than incremental progress.

  3. Competitive pressure from other labs

    New releases from OpenAI, Anthropic and xAI continuously reshape the LMArena comparison pool, since every vote is relative to whichever other models users happen to be shown. A stronger field of rivals can either pull Gemini's relative score down even if its raw ability improves, or push Google to release a markedly stronger model to keep pace.

  4. Strict labeling rule

    The settlement rules exclude Flash and Flash-Lite variants explicitly, so a strong but non-Pro release does not count. This narrows the qualifying model to whichever specific release Google designates 'Pro', adding uncertainty about exactly which model and when it will appear.

  5. Thin, single-venue liquidity

    With $70,236 traded on one venue, this market has less depth than many prediction markets on higher-profile events. That makes the 8% reading more sensitive to individual trades and less of a stable, multi-venue consensus than figures seen on larger markets.

The case for

  • Google has accelerated its Gemini release cadence over recent generations, increasing the odds a Pro-labeled model appears well before the 1 January 2027 deadline.
  • If competitive pressure from OpenAI, Anthropic or xAI forces a step-change release rather than an incremental one, LMArena voters could reward a clearly superior model with an outsized score gain.
  • A release timed early in the window would leave months of runway, reducing the timing risk that currently weighs on the price.
  • Any methodology shift at LMArena that compresses or re-scales the rating pool could make a 1500 score more reachable than historical increments suggest.

The case against

  • No Pro-labeled Google model has reached a 1500 score on debut to date, and the contract requires it on the very first measured day, not after subsequent tuning.
  • The resolution window closes 1 January 2027; if Google's next Pro release slips past that date, the contract resolves No regardless of eventual performance.
  • LMArena score gains between generations have generally been incremental rather than the scale of jump implied by crossing 1500 from below on day one.
  • A model explicitly labeled only Flash or Flash-Lite does not qualify, narrowing the set of releases that could resolve this Yes even if Google ships several models in the window.

What to watch

The clearest signal will be any official Google DeepMind announcement naming a new Gemini release 'Pro' โ€” distinct from Flash or experimental tags โ€” followed by its appearance on the LMArena Text Arena leaderboard. Once that happens, the score shown at 12:00 PM ET the following day is the number that settles this contract. Separately, watch for major releases from OpenAI, Anthropic or xAI before then, since those reshape the comparison pool Gemini will be rated against. If no qualifying Pro model has appeared by 1 January 2027, the market resolves No by default.

Trade this contract

Venues (1)

More about this event

Venues (1)

Probability

  • Will the next Google Gemini Pro model added to the Arena Leaderboard debut at a score of at least 1480?10%
  • Will the next Google Gemini Pro model added to the Arena Leaderboard debut at a score of at least 1495?8%
  • Will the next Google Gemini Pro model added to the Arena Leaderboard debut at a score of at least 1490?7%
  • Will the next Google Gemini Pro model added to the Arena Leaderboard debut at a score of at least 1500?7%
  • Will the next Google Gemini Pro model added to the Arena Leaderboard debut at a score of at least 1510?5%

Resolution rules

Determined by
lmarena.ai/leaderboard/text
Resolution date

This resolves Yes if the next Google Gemini model explicitly labeled 'Pro' (for example gemini-3-pro or gemini-3.1-pro-preview) debuts on the LMArena Text Arena Overall leaderboard, style control off, with a score of at least 1500, as shown at 12:00 PM ET the day after its first appearance. Models labeled only Flash, Flash-Lite, or other non-Pro variants do not qualify. If multiple qualifying Pro models appear on the same date, the highest-scoring one is used; if no qualifying model appears before 1 January 2027, the market resolves No. The source is lmarena.ai/leaderboard/text.

Calculation methodology โ†’

Local context

English-speaking AI and developer communities treat the LMArena leaderboard as a running scoreboard for which lab is ahead, and a high-profile Gemini debut score feeds directly into that conversation on forums, newsletters and social platforms this audience already follows. For readers tracking Alphabet as an investment or weighing which AI platform to build on, a Gemini release that visibly tops or trails rivals on LMArena is one of the more concrete, frequently cited signals used to make that judgment, even though the leaderboard measures user preference in blind comparisons rather than every dimension of real-world capability.

Common questions

What exactly settles this market, and when?
It settles based on the score shown on LMArena's Text Arena Overall leaderboard (style control off) for the next Google model labeled 'Pro', read at 12:00 PM ET the day after that model first appears. The market closes by 1 January 2027; if no qualifying model has appeared by then, it resolves No.
What does the current price actually mean?
The price reflects what traders currently think the chance of a Yes outcome is, not a guarantee. A price of 8% means the market sees this as a relatively unlikely outcome, roughly in line with happening about once in twelve similar situations, based on the information available now.
What happens if Google releases a Pro model but the score is ambiguous or the leaderboard changes its methodology?
The rules point to the score as shown at the specified 12:00 PM ET checkpoint, so whatever LMArena displays at that moment is what counts. If LMArena changes its scoring methodology before then, that would affect how comparable the new score is to past generations, though the settlement would still rely on the figure shown at the stated time.
Why does the threshold sit at exactly 1500?
LMArena uses an Elo-style rating system where 1500 is sometimes treated as a notional midpoint or reference level, but no Gemini Pro release has debuted at or above it so far. The market uses that specific figure as a clean, publicly visible bar rather than one tied to any particular prior Gemini score.
What is LMArena and why does it matter here?
LMArena has users compare anonymous pairs of chatbot responses and vote for the better one, converting those votes into a running leaderboard. It is widely cited in AI and developer communities as a rough, crowdsourced proxy for which models perform best in practice, which is why a Gemini debut score draws attention beyond the market itself.
Why is trading volume on this market so low?
At $70,236 traded on a single venue, this is a smaller market than many higher-profile prediction markets, likely reflecting that it tracks a fairly specific technical benchmark outcome with a narrow resolution window. Lower volume means the price can be more sensitive to individual trades than in deeper markets.

Related events