Scan: Will Google be the second-best AI lab at the end of September 2026?

date	predicted	market price	confidence
2026-09-02	63%	68%	50%

1. Planner Fable 5.1 · Fermi decomposition + tool plan

Tools chosen: polymarket_direct polymarket_related kalshi_related claude_news claude_news article_search gdelt_news wikipedia

Sub-questions (Fermi decomposition)

Which labs currently occupy ranks 1, 2, and 3 on the arena.ai Text Arena Overall leaderboard with style control off, and what are the score gaps between them?
How frequently has the second-place lab on the Arena leaderboard changed over the past 6-12 months, and what is the historical base rate of Google being exactly #2 vs #1 vs #3+?
What major frontier model releases (OpenAI GPT-5.x/6, Anthropic Claude, xAI Grok 5, Google Gemini 3.x, DeepSeek, Meta) are expected or rumored between now and September 30, 2026, and how have recent releases shifted Arena rankings?
What does the Polymarket 'best AI lab end of September' sibling market imply about Google's probability of being #1 (which would preclude #2)?
How do the other 'second-best lab' markets (OpenAI, Anthropic, xAI, Other) price, and do they sum consistently with the Google market?
Has arena.ai changed its lab-ranking methodology (e.g., how a lab's rank is computed from its models, style-control default) in ways that could affect resolution?

Planner reasoning

This resolves on the arena.ai Text Arena lab ranking (style control off) on Sept 30, 2026, so the key questions are who currently leads, who is second, how volatile the top-3 ordering has been, and what model releases (Gemini, GPT, Claude, Grok, DeepSeek, etc.) are expected before end of September. Google has often been #1 in Arena rankings, so the question is really whether Google gets displaced to exactly #2 rather than #1 or #3. Polymarket direct price is the primary anchor, with sibling markets (OpenAI/Anthropic/xAI second-best, and 'best lab' markets) providing consistency checks.

2. Research Parallel tool calls

tool	status	time	items	summary
polymarket_direct	OK	0.5s	1	## This Market's Polymarket Data Will Google be the second-best AI lab at the end of September 2026? - Current price (probability): 69.50% - 7-day price change: +36.00% - 30-day price change: +23.50% - Total volume: $29,353 (USD notional) - Price range: 14.00% - 82.00% - Data points: 51 days
polymarket_related	OK	4.1s	0	Scanned 100 active Polymarket markets, kept 0 matches. keyword 'second-best AI lab': 0 markets \| keyword 'best AI lab': 0 markets \| keyword 'arena.ai': 0 markets \| keyword 'LMArena': 0 markets \| keyword 'top AI model September': 0 markets
kalshi_related	OK	3.9s	1	1 related markets / summaries. keyword 'AI lab': ok \| keyword 'LMArena': no matches \| keyword 'best AI model': no matches
claude_news	OK	35.4s	14	Based on research into arena.ai (LMArena) Text Arena rankings through September 2026: Current leaderboard (Sept 2026): - Anthropic holds the #1 spot on the Text Arena overall board with Claude Fable 5. On raw human preference GPT-5.6 Sol sits #14 on Arena's text leaderboard at 1482.8, where Cl
claude_news	OK	27.2s	13	Based on research, here are key findings on the current state of frontier AI labs as of September 2026: - Anthropic currently dominates most leaderboards: On LMArena-style composite rankings, Claude Fable 5.1 leads the LLM leaderboard with an Artificial Analysis Intelligence Index of 53.4, ahe
article_search	OK	1.0s	44	Fetched 44 unique articles across 6 queries (lookback=45d). 'arena.ai text leaderboard top model': 10 articles \| 'LMArena leaderboard Gemini GPT Claude Grok ranking': 10 articles \| 'new frontier model release Arena score': 10 articles \| 'Gemini 3 arena ranking': 10 articles \| 'Grok 5 release': 10 ar
gdelt_news	FAILED	240.0s	0	timeout after 240.0s
wikipedia	OK	0.2s	4	Fetched 4 Wikipedia entries (0 missing pages).

3. Evidence Brief Sonnet 5 · 7792 chars

# Current state The market resolves off the arena.ai Text Arena Overall leaderboard (style control off) on 2026-09-30. As of Sept 2026, Anthropic (Claude Fable/Opus 5) holds a clear #1 on that board; Google (Gemini 3.1 Pro) sits in a tightly-clustered #3–5 tier along with OpenAI (GPT-5.6/GPT-6 Astra), xAI (Grok 4.5/4.6/4.1), and now Meta (Muse Spark/Astra), rather than holding a stable #2. The #2/#3 ordering among Anthropic-adjacent challengers has fluctuated monthly through 2026. # Timeline of key events - 2026-02-26 (confirmed, per claude_news citing Arena data): Opus 4.6/4.6 Thinking tied #1 (1503), Gemini 3.1 Pro Preview #3 (1500), Grok-4.20-beta1 #4 (1495). - 2026-03 (confirmed): Anthropic Elo 1503, xAI 1495, Google 1494, OpenAI 1481 — Google #3, essentially tied with xAI #2. - 2026-08 (reported): "Claude Fable 5" restored to #1 after re-baseline, ahead of a tight Opus 4.8/GPT-5.5 Pro/Gemini 3.1 Pro Preview cluster. - 2026-08-10/12 (confirmed): Meta open-sources Muse Spark/Glimmer models, entering frontier competition, per CNBC. - 2026-09-01 (reported): Analyst commentary places Meta's AI lab on the "top three podium," implying further crowding of the #2/#3 tier. - 2026-09-09 (reported): Meta's Muse Spark briefly overtook GPT-5.6 Sol on Artificial Analysis Index; OpenAI's Astra then retook second behind Anthropic's Fable. - 2026-09 (reported, felloai.com/modelgrep.com): Claude Fable 5.1 #1 on Intelligence Index (53.4), GPT-6 Astra #2 (52.8), Claude Opus 5 #3 (50.7) — Google not in top 3 of this specific ranking. - Rumored: Google skipping Gemini 3.5 Pro, targeting Gemini 4 launch ~October 2026 — after this market's close. # Event Will Google rank exactly #2 (by arena.ai Text Arena Overall, Lab Rank) among AI labs at check time on 2026-09-30? # Outcomes to forecast - Yes (Google is #2) - No (Google is not #2) # Kalshi market anchor Only a Polymarket-tagged data feed was returned for this exact ticker (0xe0441ada...), showing current price 69.5% YES, up +36pp over 7 days and +23.5pp over 30 days, range 14%–82%, volume ~$29K over 51 days. No separate Kalshi-native price was returned in raw research; treat 69.5% as the cross-platform consensus anchor, but flag that this is a thin, highly volatile market (14%→82% range) that has moved sharply recently, possibly on stale/rumor-driven momentum rather than confirmed leaderboard data. # Sub-question answers 1. **Ranks 1-3 and gaps** — Anthropic (Claude Fable/Opus) is confirmed #1 across virtually all recent snapshots. #2/#3 is contested among OpenAI (GPT-5.6/6 Astra), xAI (Grok), Google (Gemini 3.1 Pro), and now Meta (Muse Spark), with gaps often <20-30 Elo points — a near-statistical tie (claude_news, arena.ai blog). 2. **Historical base rate** — Over the 41-month Arena history since May 2023: OpenAI #1 for 15 months (37%), Anthropic 9 months, Google 8 months (claude_news). No explicit monthly #2-only breakdown was found, but Google has cycled between #2, #3, and #4 through 2026 (Feb: #3 tied w/ xAI #2; Mar: #3; Aug: in a tied cluster behind Anthropic). 3. **Upcoming releases** — Grok 5 repeatedly delayed (no confirmed date as of Aug 2026); Google reportedly skipping Gemini 3.5 Pro for Gemini 4, rumored for October 2026 (after close); OpenAI and Anthropic have shipped frequent flagship updates (GPT-5.6 Sol, GPT-6 Astra, Claude Opus 5/Fable 5.1) keeping ahead of Google's Flash-tier-only Gemini updates; Meta has newly entered with Muse Spark/Glimmer, adding competition for the #2-3 slots. 4. **Polymarket "best lab #1" implication** — Not directly retrieved (polymarket_related found 0 matching markets); claude_news references a "Best AI Model by Company" market with Anthropic as incumbent #1 favorite, implying Google is NOT priced as #1, which structurally increases (not decreases) room for Google to be #2 rather than #1. 5. **Other "second-best" markets pricing/consistency** — No data retrieved; polymarket_related and kalshi_related found no matching sibling markets for OpenAI/Anthropic/xAI second-place tickers. Cannot verify probabilities sum consistently. 6. **Methodology changes** — No explicit mention of a lab-ranking methodology change in 2026, but Arena's "Leaderboard Illusion" critique (Google/OpenAI each ~20% of battle data, private variant testing) is flagged as a persistent distortion risk (claude_news, surgehq.ai). # Key facts (high-confidence, factual) 1. [Wikipedia] Arena.ai is a public crowd-voted LLM evaluation platform used for preview releases; known limitations in methodology exist. 2. [claude_news/arena.ai] Anthropic (Claude Fable/Opus) held Text Arena #1 consistently from Feb through Sept 2026. 3. [claude_news] Google's Gemini has had no new Pro-flagship since Gemini 3.1 Pro Preview; only Flash-tier updates through Sept 2026; Gemini 4 rumored for October 2026 (post-close). 4. [CNBC, Aug-Sep 2026] Meta (Muse Spark/Glimmer/Astra) has newly entered frontier competition, further crowding the #2-4 tier. 5. [Polymarket data] Google "second-best" market priced 69.5% YES, +36pp in 7 days — sharp recent upward momentum despite research suggesting Google is not clearly #2 on the actual resolution source. # Cross-market signals - Kalshi related: No directly relevant sibling markets found (only unrelated labor-force-participation tickers matched keyword "AI lab"). - Polymarket: This exact market shows 69.5% YES with high volatility (14%-82% range) and low volume (~$29K) — thin market, could be noisy/inefficient. - Sportsbook implied: N/A. # Analyst opinions and speculation - MindStudio (qualitative, non-Arena scoring): Google rated highest on "overall lab quality" (74, tied OpenAI) but lowest on "momentum" (3/10) vs OpenAI (10/10) and Anthropic (8/10) — suggesting structural strength but stalling relative trajectory. - CNBC analyst commentary (Sept 2026): frames Meta as newly on the "top three podium," implicitly squeezing Google further down. - Surgehq.ai critique: Arena methodology may overweight certain labs; ranking volatility is partly artifact-driven. # Directional lean per outcome - **Yes (Google #2)**: Supported by Google's broad resourcing/ecosystem strength and past instances of Gemini reaching #2-3; opposed by consistent recent snapshots (Feb, March, Aug 2026) showing Google #3 or worse, stalled flagship (no new Pro model since 3.1), and now a 4th competitor (Meta) crowding the tier — net evidence leans against a clean #2 finish specifically on Text Arena. - **No (Google not #2)**: Supported by multiple independent 2026 benchmarks (Artificial Analysis, LMArena snapshots) placing OpenAI, and increasingly Meta, ahead of or tied with Google for #2/#3; Google's Gemini 4 (rumored stronger) doesn't launch until after the Sept 30 window. # Gaps / unknowns - No direct Kalshi-native order book data was retrieved (only a Polymarket-tagged feed under the same ticker) — true Kalshi consensus price is uncertain. - No sibling "second-best lab" markets (OpenAI/Anthropic/xAI/Other) were found for cross-checking implied probability consistency. - Exact live Lab Rank #2 as of the article-search window (last 45 days) is not conclusively pinned down — sources conflict on whether Google, OpenAI, or Meta currently holds #2. - Unclear why Polymarket price for a "Google #2" outcome rose sharply (+36pp in 7 days) despite research suggesting Google's Arena position is not improving — possibly rumor-driven (e.g., Gemini 4 buzz) rather than fundamentals. # Calibration anchors - Polymarket/cross-listed current price: 69.5% YES (highly volatile, thin volume). - Historical base rate: Google held Arena Text #1 in 8/41 months (~20%) historically; specific #2-only base rate not available but recent 2026 snapshots show Google oscillating between #3 and a tied #2/#3 cluster, rarely a clean, undisputed #2.

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

claude-fable-5-1

38% conf

Yes	63%
No	37%

The cross-platform anchor sits at 69.5% YES and has surged +36pp in 7 days, which close to resolution typically reflects the live Text Arena Lab Rank (i.e., Gemini currently sitting #2 behind Anthropic) rather than rumor. However, the brief's snapshots through 2026 show Google oscillating in a tightly clustered #2-#4 tier with OpenAI, xAI, and now Meta, often within 10-20 Elo of each other, so a single new OpenAI/Meta/xAI release or Arena re-baseline before Sept 30 could flip the ordering. Google has no new Pro-flagship expected before close (Gemini 4 rumored for October), so upside from Google itself is limited, while downside risk comes from competitors. I shade modestly below the market to reflect thin liquidity, high volatility (14%-82% range), and the crowded tier, but stay Yes-leaning given the market's likely informational edge on the actual leaderboard.

gpt-5.6-sol

58% conf

Yes	44%
No	56%

The 69.5% YES cross-platform price is the starting anchor, but its thin volume, extreme 14%-82% range, and unexplained 36-point weekly jump make it less reliable than a normal market consensus. Recent Arena evidence points toward No because Google has generally occupied a contested #3-5 tier, while Anthropic is securely first and OpenAI, xAI, and Meta all compete for second within small Elo gaps. Historical performance leaves Google plausible, but no pre-deadline Pro flagship is expected, whereas OpenAI and Anthropic have updated frequently and Meta has added another strong challenger. Methodological noise and the absence of a conclusive live Lab Rank prevent a stronger move away from the market anchor.

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

Both forecasts anchor heavily on the 69.5% Polymarket price, but the brief explicitly flags this as a thin (~$29K volume), highly volatile (14%-82% range) market with an unexplained +36pp jump — neither forecaster adequately discounts for the possibility this is pure noise/rumor (e.g., Gemini 4 buzz) rather than reflecting actual live Arena rank, especially since Forecast 1 explicitly assumes it reflects "the live Text Arena Lab Rank" with no direct evidence for that claim.

Neither forecast grapples with the direct contradiction between the market price (69.5% YES) and nearly all cited fundamental evidence: every confirmed 2026 snapshot (Feb, March, Aug, Sept) shows Google in a contested #3+ tier, never a clean #2, and the Sept Intelligence Index explicitly ranks Google outside top 3. This is a strong case for anchoring much further from the market than either forecast does (0.63 and 0.44 both still lean toward or near "Yes" territory relative to the fundamentals).

Both underweight the base-rate evidence: "Google has cycled between #2, #3, and #4 through 2026" with no month cited where Google held an undisputed #2 — this suggests the "No" case may deserve more probability mass than either forecast assigns, particularly Forecast 1's 63% Yes.

Neither forecaster addresses Meta's entry as a structural, not just marginal, threat — Meta going from 0 to "top three podium" commentary in a single month (Aug-Sept 2026) is a fast-moving new source of downside risk for Google's #2 chances that could easily persist or worsen by Sept 30, yet both treat it as just one more name in the cluster rather than actively increasing "No" probability over time.

Both forecasts fail to consider the resolution's brittleness: since the market resolves on a single snapshot read of Arena rank at a specific check time (2026-09-30), and the tier is "near-statistical tie" with gaps <20-30 Elo, this argues for a wider variance / closer-to-50% outcome (reflecting genuine coin-flip uncertainty among 3-4 clustered labs) rather than the confident directional bets both forecasts make (63/37 and 44/56).

Neither addresses the Arena methodology risk explicitly flagged in the brief (Leaderboard Illusion critique — "Google/OpenAI each ~20% of battle data, private variant testing") which could artificially inflate or deflate Google's rank independent of true model quality, adding a source of resolution-criteria uncertainty that should widen forecaster uncertainty bands rather than support confident point estimates.

6. Aggregator Fable 5.1 · final calibrated forecast

57% (48% confidence)

Yes: 57% No: 43%

The resolution source is the arena.ai Text Arena Lab Rank on a single snapshot, and the brief never pins down the live #2, so the sharp +36pp move to 69.5% YES in the final week is the most direct (if thin) signal that Gemini currently sits #2 behind Anthropic — Gemini models historically over-index on Text Arena relative to Intelligence Index-style benchmarks, so the Sept Artificial Analysis ranking that excludes Google is weak counter-evidence for this specific board. However, every confirmed 2026 Arena snapshot shows Google in a near-statistical-tie cluster (#2-#4 within ~10-20 Elo) with OpenAI, xAI, and now Meta, and Google has no Pro-flagship before close while competitors ship frequently, so any late release or re-baseline could flip the ordering. The critique is right that both forecasters lean too hard on a thin, volatile market, but it overstates the case for No: the base-rate objection concerns 'undisputed' #2, whereas the market only requires Google to hold the #2 slot at check time, which is roughly a 3-way coin flip tilted toward Google if the recent price move reflects the live board. I settle between the two forecasts, shading below the market for liquidity and cluster-flip risk but staying Yes-leaning.

Pipeline Timing

Total pipeline time: 346.5s

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