Gemini 3.5 Pro debut arena score? — above 1515

- [2026-09-30] Google DeepMind launched "Gemini 4 Argon" (not labeled "Gemini 3.5 Pro") — the first proprietary model above the Flash line since Gemini 3.1 Pro, confirmed via official Arena.ai account. "Introducing Gemini 4 Argon – our new frontier model. It's built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program." - [2026-09-30] Arena.ai (the resolution source) reported Gemini 4 Argon (High) debuted at #1 on Text Arena with a score of 1525, based on 4,942 votes. "Big news: Gemini 4 Argon (High) by @GoogleDeepMind just landed #1 in Text Arena with 1525 pts, and #8 in Code Arena: WebDev with 1679 pts!" This is well above the 1515 strike level. - [2026-10-02] Independent reporting corroborates the 1525 figure and places it ~20 points ahead of the #2 model. "Argon leads the LMArena Text Arena at 1,525 points, 20 points ahead of Claude Opus 4.6 (High) in second." - [2026-10-01] Access to Gemini 4 Argon remains extremely limited: it is rolling out only to "government users and trusted cyber defenders" "Availability starts with government users and trusted cyber defenders in the Fairwind Program. Google says it will refine guardrails before opening access to developers, enterprises and consumers." - [2026-09-23] As of late September, Gemini 3.5 Pro itself had still not shipped at all. "Google DeepMind still lists Gemini 3.5 Pro as coming soon, and Google has not announced a release date." - [2026-08-23] Gemini 3.5 Pro had already missed three announced/rumored windows (June, mid-July, early August) and had no model ID or pricing. "Gemini 3.5 Pro is still unreleased as of 23 August 2026... It has now missed three targets (June, mid-July, early August), and has no model ID, no pricing and no launch date." - [2026-08-14] A separate forecasting tracker placed the median expectation for Gemini 3.5 Pro's general-availability release at October 31, 2026, with wide uncertainty extending into 2027. "An early fall release remains plausible if partner testing concludes successfully, but the median expectation falls in late October, set at October 31... As of August 14, 2026, Google has not released Gemini 3.5 Pro." - [2026-02-19] For context on the pre-Argon ceiling: Gemini 3.1 Pro (the last numbered "Pro" release before Argon) debuted around 1500 on Text Arena, well below 1515. "It's top 3 across Text and Vision Arena, and #6 in Code Arena, tied closely with Claude Opus 4.5. Highlights: ▪️ Tied #1 in Text (scoring 1500), 4 pts from Opus 4.6" - [2025-11-18] Gemini 3 Pro's original debut score was 1501 on Text Arena. "Notably, Gemini 3 Pro tops the LMArena leaderboard with a score of 1,501 points, dethroning Grok 4.1 Thinking, which earned a score of 1,483 points." - [2026-10-03] A Gemini-labeled model (gemini-4-argon-high) is now shown sitting atop the official Text Arena leaderboard snapshot at a rating around 1525-1533 depending on snapshot date, versus Claude-family models clustered at 1498-1512. BOTTOM LINE: Gemini 3.5 Pro (the specifically named model) has still not been released as of early October 2026 and remains in "coming soon" status with no firm date, making a debut score under that exact name unlikely before the Jan 1, 2027 deadline. However, Google has instead jumped straight to a new flagship, "Gemini 4 Argon," which debuted Sep 30, 2026 with a Text Arena score of 1525 — above the 1515 threshold — raising a genuine adjudication question of whether "Gemini 4" satisfies the rule's "Gemini 3.5 Pro or greater" labeling clause, since it is a numerically later model but does not carry the "3.5 Pro" or even "Pro" label. NOT FOUND: No official LMArena/Arena.ai leaderboard page snapshot showing the exact "Remove Style Control" (style-control-off) Text Arena number for Gemini 4 Argon or Gemini 3.5 Pro was accessible directly (the 1525/1533 figures found come from the default/no-style-control-stated Arena posts and third-party trackers, some of which may be AI-generated aggregator sites rather than the primary lmarena.ai site itself, which could not be fetched directly). No explicit Kalshi or Arena rules-committee statement was found clarifying whether "Gemini 4" qualifies as "Gemini 3.5 Pro or greater" for this specific contract.

- [2025-11-18] Gemini 3 Pro debuted on LMArena Text at 1501 Elo, becoming #1 at launch, a "breakthrough score" per Google's own announcement — the only prior Gemini "major version" debut score available as reference. https://blog.google/products/gemini/gemini-3/ - [2026-02-19] Gemini 3.1 Pro followed roughly three months later; third-party trackers reported its debut Arena score in the 1486–1500 range (one tracker: "Tied #1 in Text (scoring 1500), 4 pts from Opus 4.6"), i.e. a flat-to-modest gain over Gemini 3 Pro, not a large jump. https://x.com/arena/status/2024519891295089063 - [2026-07-31] Multiple unverified leak reports claimed an identifier "gemini-3.5-pro" briefly appeared in LMArena blind testing for under an hour with "no official benchmark rankings or Elo scores" disclosed, and independent reviewers found no preserved Arena result or screenshot to corroborate the leak. https://windowsforum.com/windows-news.4/gemini-3-5-pro-no-launch-date-or-verified-lm-arena-test.441354/ - [2026-09-23] As of late September 2026, Google DeepMind's own model page still lists Gemini 3.5 Pro as "coming soon," with Bloomberg reporting the flagship is "months behind schedule, held up by coding performance." https://aiwiz.uk/blog/gemini-3-5-pro/ - [2026-09-09] An independent tracker states "the only official wording is that it is testing with partners and will be broadly available when ready — no date has ever been given," confirming no public release or Arena debut score exists yet. https://qcode.cc/en/gemini-3-5-pro-guide - [2026-10-01] By October 2026, general LMArena leaderboard trackers show the top text-arena scores clustering around 1500–1525 (e.g., Claude-family models cited near 1509–1525, Gemini 3.1 Pro Preview around 1486–1500), indicating the "above 1515" bar sits at or slightly above the current overall frontier ceiling, not just above Gemini's own prior scores. https://localaimaster.com/blog/lmarena-chatbot-arena-leaderboard - [2026-10-01] The historical pattern for the past two Gemini "dot" releases (3→3.1) shows debut Arena scores essentially flat or marginally improved (1501→~1486-1500) rather than reliably climbing by double digits, which is the scale of jump (vs. 1501) needed to clear 1515. https://x.com/arena/status/2024519891295089063 - [2026-10-03] No model has yet been officially labeled "Gemini 3.5 Pro" or higher on any LMArena/Arena.ai leaderboard; all available data points are rumor/leak-tier, not resolution-grade sources. https://windowsforum.com/windows-news.4/gemini-3-5-pro-no-launch-date-or-verified-lm-arena-test.441354/ BOTTOM LINE: The reference class (Gemini 3 Pro → 3.1 Pro) shows debut Arena scores historically landing in the 1486–1501 range with only modest increments between generations, and the current overall Arena frontier ceiling sits around 1500–1525 — so clearing 1515 would require a larger-than-typical generational jump; moreover, as of early October 2026 Gemini 3.5 Pro has not even launched or appeared with a confirmed Arena score, adding substantial timeline risk before the January 1, 2027 deadline. NOT FOUND: No official Google or LMArena/Arena.ai confirmation of a Gemini 3.5 Pro release date, model card, or any verified debut Elo score; all "leak" reports of its appearance on Arena are explicitly unverified by independent reviewers.

YES means: a Gemini model officially labeled "Gemini 3.5 Pro" or a higher/later-numbered Gemini model (per the rule "Gemini 3.5 Pro or greater") is added to the Arena Text leaderboard (Remove Style Control) and posts a debut score above 1515, before Jan 1, 2027. - [2026-09-30] Google DeepMind's own announcement states Gemini 4 Argon directly supersedes the Pro line: 9to5Google's launch-day report says Argon It follows the cancellation of Gemini 3.5 Pro and focus on 3.8 Flash. This is Google's clearest acknowledgment that 3.5 Pro was abandoned in favor of a numerically "greater" model, strengthening the case that Argon itself satisfies the rule's "Gemini 3.5 Pro or greater" labeling clause even without carrying the exact "3.5 Pro" name. - [2026-09-30] Independent outlet The New Stack frames Argon explicitly as the replacement for the promised Pro model: "Google first announced plans for a new Pro model, Gemini 3.5 Pro, at its I/O developer conference in May — Argon is essentially its replacement." - [2026-10-01] Engadget likewise reports "It seems the company has moved beyond Gemini 3.5 Pro," and that Argon "follows Gemini 3.5," reinforcing that Argon is the de facto successor model in the version sequence that the Kalshi rule intends to cover. - [2026-09-30] Arena.ai's own X account posted the debut result directly: Big news: Gemini 4 Argon (High) by @GoogleDeepMind just landed #1 in Text Arena with 1525 pts, and #8 in Code Arena: WebDev with 1679 pts! This release has reshaped the Text Arena Pareto frontier with a blended $8/MToken! This 1525 figure, from the primary Arena source itself, clears the 1515 strike by 10 points even before any style-control adjustment. - [2026-10-01] therundown.ai, citing Arena directly, confirms the debut context and caveats: Argon's high setting took first place on Arena's September 30 text leaderboard, based on 4,942 votes. Arena marked the result preliminary. A "preliminary" label does not retract the number; historically preliminary Arena scores solidify rather than reverse once vote counts grow. - [2026-09-30] Arena's changelog (arena.ai/company/leaderboard-changelog) independently corroborates the addition: gemini-4-argon-high has been added to Code Arena: WebDev, Agent Arena, and Text Arena leaderboards. This is a primary-source confirmation (not a third-party aggregator) that the model is now formally on the resolution venue's leaderboard under the "gemini-4-argon-high" label, satisfying the "added to the leaderboard" prerequisite in the rules. - [2026-10-01] A dedicated benchmark tracker (swen.live) lists the raw Elo as 1525.0, matching Arena's own figure and providing a second independent corroboration outside of X/Twitter screenshots. - [2026-10-01] felloai.com's write-up, aggregating Arena's category breakdowns, notes Argon also topped sub-categories: It also ranked first in the coding, hard prompts, instruction following and creative writing categories. This breadth suggests the 1525 score is not an artifact of a narrow category but a broad, robust win margin (reported ~20 points over the #2 model), giving headroom above 1515 even if the score drifts down somewhat once more votes are collected or style control is applied. - [2026-10-04] No source found yet shows the exact "Remove Style Control" variant score for Argon, but historically Gemini models' style-controlled scores have tracked closely with (within a few points of) their raw scores, so a ~20-point buffer above the 1515 strike provides meaningful cushion against any style-control discount. - [2026-09-30] Google's access-gating (Fairwind Program only, per the official blog language rolling out to "trusted testers") does not bar Arena qualification: the model is already live on the Arena leaderboard with thousands of votes, meaning limited public availability is not an obstacle to resolution, only to broader adoption. BOTTOM LINE: The single most likely path to YES is an adjudicator reading "Gemini 3.5 Pro or greater" to include Gemini 4 Argon, Google's own replacement for the cancelled 3.5 Pro, which already sits on Arena's official Text leaderboard at a primary-source-confirmed 1525 points — ten points clear of the 1515 strike — well before the Jan 1, 2027 deadline. NOT FOUND: An explicit, dated Kalshi/Arena rules clarification stating that "Gemini 4 Argon" (or any model not literally labeled "Gemini 3.5 Pro") counts as satisfying the "3.5 Pro or greater" clause; and the exact lmarena.ai "Remove Style Control" numeric score for Argon (all figures found are default/no-style-control-stated Arena posts and secondary trackers).

YES means: a model officially named "Gemini 3.5 Pro" (or a later/greater labeled Gemini model) is added to the Arena AI Text Score leaderboard with a Remove-Style-Control score above 1515 before Jan 1, 2027. - [2026-10-01] Multiple independent outlets report Google has effectively abandoned the "Gemini 3.5 Pro" name: Digitimes states Google "unveiled Gemini 4 Argon...after months of delays and the cancellation of Gemini 3.5 Pro," meaning the exact named model the contract targets may never exist at all. (https://www.digitimes.com/news/a20261001VL209/google-gemini-openai-software-anthropic.html) - [2026-09-30] 9to5Google confirms the same framing directly from launch coverage: Gemini 4 Argon "follows the cancellation of Gemini 3.5 Pro and focus on 3.8 Flash," a mechanism rather than rumor at this point since it is now the consensus narrative across outlets. (https://9to5google.com/2026/09/30/gemini-4-argon-announcement/) - [2026-10-01] One outlet notes the unresolved adjudication risk explicitly: "Google has not confirmed that Argon is simply 3.5 Pro under a new name," and it "did not announce that the model had been renamed" — so there is no official document equating Argon with "Gemini 3.5 Pro or greater" in Google's own naming scheme, leaving the Kalshi qualifying clause resting on inference, not confirmation. (https://meetcody.ai/blog/gemini-4-argon-gemini-pro-delay/) - [2026-10-02] A sibling Polymarket market's own resolution notes state plainly that qualifying requires a model "explicitly labeled with the relevant version name," and that "labeling errors, placeholder text, or version names...that do not correspond to a model that is actually accessible...will not qualify" — showing other venues treat exact-label matching as a strict, literal test, which would cut against counting "Argon" (no "Pro," no "3.5," no explicit "4" version label in the product name) as satisfying "Gemini 3.5 Pro or greater." (https://polymarket.com/event/gemini-4pt0-released-by-june-30-2026) - [2026-10-02] The live Arena Text leaderboard snapshot lists "gemini-4-argon-high" at 1525±9 points and explicitly flags the result "Preliminary" on only 4,932 votes, meaning the published score carries a stated margin of error whose lower bound (1516) sits barely one point above the 1515 strike, leaving real statistical risk the finalized (non-preliminary) rating could still settle at or below 1515 once more votes accrue. (https://arena.ai/leaderboard/text/) - [2026-10-02] That same leaderboard view is displayed with the "Style Control" adjustment toggle set to "None," meaning no independently confirmed, explicitly labeled "Remove Style Control" figure distinct from the headline 1525 was located; if Kalshi's resolution source requires a specifically toggled style-control-off figure that diverges from the default/preliminary number reported in press coverage, no primary citation pinning that exact figure was found. (https://arena.ai/leaderboard/text/) - [2026-09-23] Prior reporting already on file confirms Gemini 3.5 Pro itself had not shipped and had no model ID as of late September 2026 — status quo inertia argues that absent an explicit Google re-christening of a future "Pro"-branded release, the named qualifying event (an official "Gemini 3.5 Pro" launch) is off the table entirely for the remainder of the contract window. - [2026-02-19] The historical base rate (Gemini 3 Pro → 3.1 Pro, debut scores 1501 and ~1487-1500) shows Google's own "Pro"-labeled debuts have never cleared 1515 on Arena Text prior to Argon, so any strict reading requiring a "Pro"-labeled release to hit the bar has a 0-for-2 track record to date. - [2026-09-30] Google's own Fairwind Program rollout notes restrict Argon initially to "government users and trusted cyber defenders," with broader access pending further safety work — if Kalshi's resolution requires the qualifying model to be the one generally reflected on a stable, non-preliminary leaderboard entry (rather than a narrow early-access cohort's preliminary vote sample), the thin, non-general-public vote base is a further technical vulnerability for the score being revised down or disputed before Jan 1, 2027. BOTTOM LINE: The strongest NO case rests on two independent technicalities: (1) the named model "Gemini 3.5 Pro" has reportedly been cancelled and will likely never launch, and Google has not confirmed "Gemini 4 Argon" as its labeled successor under the "3.5 Pro or greater" naming the contract requires; and (2) even taking Argon's reported 1525 Text Arena score at face value, it is explicitly marked "Preliminary" with a ±9 margin whose floor sits only 1 point above the 1515 strike, leaving room for the finalized/style-control-adjusted figure to land at or below threshold. NOT FOUND: No official Kalshi/market-sponsor adjudication statement on whether "Gemini 4 Argon" counts as "Gemini 3.5 Pro or greater"; no primary lmarena/Arena.ai page displaying a distinctly labeled "Remove Style Control" number for Argon that differs from the widely cited 1525 preliminary figure; no official Google confirmation (only third-party/analyst reporting, e.g., SemiAnalysis-sourced) that Gemini 3.5 Pro was formally cancelled.