Scan: Will the next Google Gemini Pro model added to the Arena Leaderboard debut at a score of at least 1510?

date	predicted	market price	confidence
2026-09-03	48%	60%	52%
2026-08-11	17%	17%	50%

1. Planner Fable 5.1 · Fermi decomposition + tool plan

Tools chosen: polymarket_direct polymarket_related kalshi_related article_search claude_news gdelt_news wikipedia code_execution

Sub-questions (Fermi decomposition)

What is the current highest score on the Arena.AI Text Arena Overall leaderboard with style control OFF, and what did gemini-3-pro (and gemini-2.5-pro) score at their debuts on that same setting?
Which Gemini Pro model is expected next (e.g., gemini-3.1-pro-preview, gemini-3.5-pro), has it been announced or leaked, and what is its expected release timing before Dec 31, 2026?
How have recent frontier debuts (GPT-5.x, Claude Opus 4.5/4.6, Grok 4.x) scored on debut with style control off, and has the top of the leaderboard moved above ~1500 recently?
How large is the gap between style-control-on and style-control-off scores for top Gemini models, and does the leaderboard default/score column show the same numbers?
What implied probability distribution over debut scores do the sibling Polymarket threshold markets (e.g., 1490, 1500, 1520, 1530) show, and is the 1510 price consistent with them?
Have Google's recent models tended to debut at #1 on Arena by a meaningful margin, and has Google shifted toward releasing 'Pro' models under codenames (anonymous testing) before appearing labeled?

Planner reasoning

This is a threshold question on the debut Arena score of the next Gemini Pro model. The key drivers are: the current top scores on the Text Arena leaderboard (style control off), how prior Gemini Pro debuts scored relative to the frontier at the time, what model is expected next (gemini-3.1-pro etc.) and its reported benchmark performance, and how the Arena Elo scale has been drifting. The Polymarket price and sibling threshold markets (e.g. 1490/1500/1520) give the implied distribution to anchor against.

2. Research Parallel tool calls

tool	status	time	items	summary
polymarket_direct	OK	0.5s	1	## This Market's Polymarket Data Will the next Google Gemini Pro model added to the Arena Leaderboard debut at a score of at least 1510? - Current price (probability): 23.70% - 7-day price change: -4.65% - 30-day price change: +7.60% - Total volume: $41,072 (USD notional) - Price range: 4.10% -
polymarket_related	OK	17.1s	1	Scanned 100 active Polymarket markets, kept 1 matches. keyword 'Gemini Pro Arena': 0 markets \| keyword 'Arena Leaderboard debut': 0 markets \| keyword 'Gemini 3.1': 0 markets \| keyword 'LMArena': 1 markets
kalshi_related	OK	17.0s	0	0 related markets / summaries. keyword 'Gemini': no matches \| keyword 'LMArena': no matches \| keyword 'Arena leaderboard': no matches
article_search	OK	36.0s	27	Fetched 27 unique articles across 4 queries (lookback=60d). 'Gemini 3.1 Pro release Arena leaderboard score': 10 articles \| 'LMArena text leaderboard top score Gemini 3 Pro 1501': 10 articles \| 'Google next Gemini Pro model release date': 10 articles \| 'Arena leaderboard style control off score fron
claude_news	OK	42.6s	18	Based on research (noting these reflect the simulated forward timeline through September 2026): - Gemini 3 Pro debut (Nov 18, 2025): Gemini 3 Pro debuted at the top of the LMArena text-reasoning leaderboard, posting a preliminary Elo score of 1501 based on pre-release community voting — the fi
gdelt_news	OK	214.4s	0	GDELT: 0 articles across 3 queries (lookback=45d). 'Gemini 3.1 Pro': error GDELT rate-limited after retries (429) \| 'LMArena leaderboard Gemini': error GDELT rate-limited after retries (429) \| 'Google Gemini Pro new model': error GDELT rate-limited after retries (429)
wikipedia	OK	0.2s	2	Fetched 2 Wikipedia entries (0 missing pages).
code_execution	OK	45.4s	0	## Key Findings - Historical debut increments (Gemini Pro/flagship vs. prior Arena #1, style‑control off): −11, −21, +28, +32, +19 → recent generations (2.0 Pro, 2.5 Pro releases) have consistently overtaken the prior leader by ~19–32 pts, while earlier 1.5 Pro releases debuted below the le

3. Evidence Brief Sonnet 5 · 6123 chars

# Event Will the next Google Gemini Pro model added to the Arena.ai Text leaderboard (style control OFF) debut with a score ≥1510? # Outcomes to forecast - Yes (debut score ≥1510) - No (debut score <1510) # Kalshi market anchor No kalshi_direct data was returned for this ticker. The only direct market pricing available is from polymarket_direct on the identical ticker: **current YES price 23.7%**, down 4.65% over 7 days but up 7.6% over 30 days, range 4.1%–51% over 91 days, volume ~$41k. Treat this as the consensus anchor in lieu of Kalshi data. # Sub-question answers 1. **Current top score / prior Gemini debut scores** — Claude News synthesis: leaderboard "Max score" was 1508 as of 2026-09-13 (Claude Fable 5, ~1508.6, #1). Gemini 3 Pro debuted at 1501 (Nov 2025, first model to cross 1500). Gemini 3.1 Pro debuted tied #1 at 1500 (Feb 2026), ~4 pts behind Opus 4.6. 2. **Next Gemini Pro model** — Gemini 3.5 Pro, announced at I/O May 2026 (June target), repeatedly delayed; still in internal/partner testing as of mid-Sept 2026 per CNBC (2026-07-21/22). Google is also already pretraining Gemini 4. No confirmed Arena appearance yet; only unverified "stealth"/codename ("Cappuccino") rumors (article_search, claude_news). 3. **Recent frontier debuts/leaderboard ceiling** — Current frontier cluster sits ~1500–1510+: Claude Opus 4.8 and GPT-5.5 Pro ~1510, Claude Fable 5 ~1508.6 (#1), Gemini 3.1 Pro ~1500, slightly behind cluster (claude_news). Ceiling has crept up gradually (1501→1500→~1508) rather than jumping decisively past 1510. 4. **Style-control on/off gap** — Not directly quantified in research; resolution explicitly uses "style control OFF" column, and reported scores above (1501, 1500, 1508.6) are stated as being from/comparable to that setting per claude_news citations of the no-style-control leaderboard URL. 5. **Sibling market distribution** — Code-execution tool modeled illustrative sibling thresholds (1490/1500/1510/1520) implying a smooth decreasing CDF (0.79/0.65/0.49/0.33); actual Polymarket price for 1510 (this market) is 23.7%, notably more bearish than the model's 49% fit — suggesting market participants weight delay/uncertainty risk and generational-flatness (3→3.1 was ~flat) more heavily than a naive Gaussian fit. 6. **Google debut margin / stealth-testing pattern** — Historical increments (code_execution): Gemini 1.5-era debuts were -11/-21 pts vs prior leader; 2.0/2.5-era debuts were +19/+28/+32 pts. Recent Gemini 3→3.1 was roughly flat (tied #1, not a fresh leap). Google has previously used stealth Arena testing for other models (e.g., "Nano Banana" per Wikipedia), and rumors suggest similar stealth testing for Gemini 3.5 Pro, but nothing confirmed. # Key facts (high-confidence, factual) 1. [claude_news] Gemini 3 Pro debuted 2025-11-18 at 1501 (first model >1500). 2. [claude_news] Gemini 3.1 Pro debuted ~Feb 2026 at 1500, tied #1, 4 pts behind Opus 4.6. 3. [claude_news] Leaderboard max score ≈1508 (Claude Fable 5) as of 2026-09-13. 4. [CNBC 2026-07-21/22] Gemini 3.5 Pro missed its June 2026 target; still in partner testing as of late July 2026; Google is investing in Gemini 4 concurrently. 5. [polymarket_direct] This exact market prices YES at 23.7%, down from a 30-day high near 51% low volatility band, current 30d trend +7.6%. 6. [polymarket_related] A separate market "Will any AI model reach 1520 by Sep 30 2026" prices YES at 99.95%, implying some (non-Gemini) model is expected to/has already cleared 1520 — showing the overall ceiling is above 1510-1520 even if Gemini specifically lags. # Cross-market signals - Kalshi related: none found (0 matches). - Polymarket (same ticker): 23.7% YES, declining short-term. - Polymarket related: "any AI model reach 1520 by Sep 2026" near-certain YES (99.95%) — indicates non-Gemini frontier models (Opus, GPT-5.5 Pro) are already near/above the 1510-1520 range, raising the bar Gemini 3.5 Pro must clear. - No sportsbook-style signals available. # Analyst opinions and speculation - code_execution model: fitted Gaussian on historical debut increments centers at ~1509 (assuming current top ~1500), giving P(≥1510)≈49%, but flags high sensitivity to the "current top score" input (33%-65% swing). - claude_news synthesis: notes Gemini 3→3.1 was essentially flat/tied rather than a clear leap, arguing recent Gemini Pro releases have not shown the "new generation" +19/+32 pt jumps seen in 2.0/2.5 eras — a bearish signal for a decisive ≥1510 debut. - Rumored Gemini 3.5 Pro specs (2M context, "Deep Think" layer, stronger agentic capability) suggest potential for a bigger leap if/when released, but unverified and pending repeated delays. # Directional lean per outcome - **Yes (≥1510):** Supported by rumored major capability upgrade in Gemini 3.5 Pro (2M context, Deep Think), and precedent of large generational jumps (2.0/2.5 eras +19–32 pts); current leaderboard ceiling (~1508) is very close to threshold, so even a modest generational bump could clear it. - **No (<1510):** Supported by recent Gemini 3→3.1 debut being flat (~1500-1501, not a leap), repeated release delays reducing near-term likelihood/urgency of an ambitious debut, market pricing (23.7%) notably below naive statistical fit (49%), and competitor models (Opus, GPT-5.5 Pro) already occupying the 1508-1510+ tier, raising competitive bar. # Gaps / unknowns - No confirmed Gemini 3.5 Pro release date or Arena debut; timing before Dec 31, 2026 uncertain given ongoing delays. - No official kalshi_direct price was retrieved — anchor relies on Polymarket pricing for identical ticker. - Precise style-control-on vs off numerical gap not quantified in sources. - No confirmation whether "next Gemini Pro" could instead be an accelerated Gemini 4 Pro variant or a smaller interim Pro release. # Calibration anchors - Polymarket YES price (same ticker, primary anchor): 23.7% - Model-implied fair value (code_execution Gaussian fit): ~45-55%, highly sensitive to current top-score assumption - Precedent: Gemini 3→3.1 Pro debut was flat (~1500-1501); Gemini 2.0→2.5 debuts leapt ahead by 19-32 pts — mixed reference class.

4. Forecasters Fable 5.1 + GPT-5.6 Sol · in parallel

claude-fable-5-1

42% conf

Yes	28%
No	72%

The only direct market pricing (Polymarket, identical ticker) sits at 23.7% YES, and the brief surfaces no confirmed Gemini 3.5 Pro release or Arena appearance despite repeated delays, so timing risk alone weighs heavily on No. The frontier ceiling (~1508 for Claude Fable 5, ~1510 for Opus 4.8/GPT-5.5 Pro) is close to the 1510 threshold, and a true new-generation Gemini flagship historically debuts at or near #1, which gives Yes a real path if it ships at all. However, the most recent Gemini 3→3.1 debut was flat (~1500-1501), Arena score gains have compressed to a few points per release, and the style-control-OFF column adds unquantified noise. I lean modestly above the market to reflect the Gaussian-fit signal and Google's incentive to debut at #1, but stay well below 50% given release uncertainty and score compression.

gpt-5.6-sol

58% conf

Yes	25%
No	75%

The 23.7% Polymarket YES price on the identical ticker is the best available consensus anchor, and I move only slightly above it. A score of 1510 is plausible because the current leaderboard ceiling is already about 1508.6 and prior major Gemini generations achieved 19–32 point gains. However, Gemini 3 and 3.1 Pro debuted essentially flat around 1500–1501, while Gemini 3.5 Pro has faced repeated delays and its rumored improvements remain unverified. The unquantified style-control effect and uncertainty over which Pro model appears next further favor No.

5. Devil's Advocate Sonnet 5 · red-team critique

Neither forecaster explicitly addresses the "will it even debut by close date" timing risk as a distinct sub-question from "if it debuts, will it hit ≥1510" — the brief flags Gemini 3.5 Pro is still in partner testing as of July 2026 with repeated delays and no confirmed release date, which should push No probability higher via a compound-probability framing (P(debuts in time) × P(score≥1510 | debuts)) rather than treating it as one blended judgment call.

Both forecasts land at 25-28% Yes, essentially anchoring tightly to the Polymarket 23.7% price without meaningfully engaging with the code_execution Gaussian fit (~49%) or explaining why that model should be discounted so heavily beyond "market is bearish" — this is somewhat circular reasoning (deferring to market because market says so) rather than independently reconciling the two divergent estimates.

Neither forecast discusses the gap/unknown that "next Gemini Pro" could be a Gemini 4 Pro variant or interim release rather than "Gemini 3.5 Pro" as assumed — since Google is "already pretraining Gemini 4" concurrently, a leapfrog release skipping 3.5 could plausibly score higher (new generation effect) and change the reference class entirely, which neither rationale considers.

The related Polymarket signal ("any AI model reach 1520 by Sep 2026" at 99.95% YES) implies competitors already sit near/above 1520, raising the competitive bar Gemini must clear to even be "at #1" — neither forecaster incorporates this specific data point into their reasoning about how much higher than 1508 the next debut would need to be to look credible/newsworthy, which is a missed calibration anchor.

Both forecasts cite "style-control OFF adds unquantified noise" as a reason favoring No, but this is asymmetric reasoning — unquantified noise could just as easily push the debut score up rather than down, and neither forecaster justifies why the noise should be treated as a bearish factor rather than a symmetric uncertainty widener.

Neither addresses the 30-day price trend (up from a low near 4.1%, +7.6% over 30 days despite a recent -4.65% weekly dip) — this directional information (recent upward momentum before a short pullback) could suggest emerging positive news (e.g., stealth testing progress) that neither forecaster incorporates into a forward-looking adjustment.

6. Aggregator Fable 5.1 · final calibrated forecast

27% (50% confidence)

Yes: 27% No: 73%

The Polymarket price on the identical ticker (23.7%) is the primary anchor, and both forecasters landed just above it; the critique cuts both ways (timing/delay risk pushes No higher, while the Gaussian fit and the fact that competitors are already at ~1510–1520 mean a Gemini flagship debuting near #1 would clear 1510 fairly easily). Conditional on a new Gemini Pro actually landing on the leaderboard before resolution, ≥1510 is roughly a coin flip or better given the frontier ceiling has moved past the threshold, but the repeated Gemini 3.5 Pro delays, unconfirmed release timing, and the flat 3→3.1 precedent substantially discount that path. The style-control noise is symmetric and should not be treated as bearish, which is a small reason not to go below market. Netting these, a modest premium to market is warranted, landing at 27% Yes.

Pipeline Timing

Total pipeline time: 329.5s

Per-tool research timings shown in the Research section above.