Scan: Will Google have the best Text-to-Video AI at the end of October 2026?

date	predicted	market price	confidence
2026-09-01	62%	68%	40%
2026-08-25	57%	58%	48%

1. Planner Fable 5.1 · Fermi decomposition + tool plan

Tools chosen: polymarket_direct polymarket_related kalshi_related claude_news article_search gdelt_news wikipedia code_execution

Sub-questions (Fermi decomposition)

Which company's model currently holds rank 1 on the arena.ai Text-to-Video Arena leaderboard (excluding AutoEval-marked models), and what is the score gap to #2?
How often has the #1 position on the arena.ai (LMArena) text-to-video leaderboard changed hands over the past 12 months, and how long has each leader typically held the top spot?
Has Google announced or is it expected to release a new Veo version (e.g., Veo 4) before end of October 2026, and how did prior Veo releases affect leaderboard rank?
Are competitors (OpenAI Sora 3, Kuaishou Kling 3.x, ByteDance Seedance 2, MiniMax Hailuo, Alibaba Wan, Runway Gen-5, xAI Grok Imagine) expected to ship new text-to-video models in the next few months that could overtake Google?
What implied probabilities do the sibling Polymarket markets in this event group assign to each company, and do they sum to ~100% after normalization?
Does Google have multiple models near the top of the leaderboard (increasing robustness of its rank-1 claim) versus single-model reliance by competitors?

Planner reasoning

The question hinges on the arena.ai Text-to-Video leaderboard state on Oct 31, 2026, which depends on Google's current position (Veo series), competitor releases (OpenAI Sora, Kling/Kuaishou, ByteDance Seedance, MiniMax Hailuo, Runway, Alibaba Wan, xAI), and how sticky rank-1 positions have historically been. The Polymarket price is the primary anchor; sibling markets in the same event group give implied probabilities for each company which should be normalized.

2. Research Parallel tool calls

tool	status	time	items	summary
polymarket_direct	OK	0.4s	1	## This Market's Polymarket Data Will Google have the best Text-to-Video AI at the end of October 2026? - Current price (probability): 65.50% - 7-day price change: +0.50% - 30-day price change: +21.00% - Total volume: $25,437 (USD notional) - Price range: 22.00% - 74.50% - Data points: 33 days
polymarket_related	OK	1.9s	0	Scanned 100 active Polymarket markets, kept 0 matches. keyword 'text-to-video': 0 markets \| keyword 'best text-to-video AI': 0 markets \| keyword 'Veo': 0 markets \| keyword 'video AI end of October': 0 markets
kalshi_related	OK	1.8s	1	1 related markets / summaries. keyword 'text-to-video': no matches \| keyword 'video AI': ok \| keyword 'Veo': no matches
claude_news	OK	48.0s	16	Here are the key findings from research on the Text-to-Video Arena landscape: Arena.ai / LMArena Leaderboard status - A snapshot from arena.ai dated 2026-03-21 shows rank 1 as "veo-3.1-audio-1080p" from Google with a score of 1381, ci ±8, and 5537 votes — Google held the top spot for much of
article_search	OK	46.8s	39	Fetched 39 unique articles across 5 queries (lookback=60d). 'arena.ai text-to-video leaderboard rank 1': 10 articles \| 'Google Veo new version release': 10 articles \| 'Sora OpenAI video model leaderboard': 10 articles \| 'Kling Seedance Hailuo text-to-video benchmark top': 10 articles \| 'LMArena vide
gdelt_news	OK	106.9s	20	GDELT: 20 articles across 3 queries (lookback=45d). 'Veo text-to-video leaderboard': 10 hits \| 'text-to-video arena leaderboard': 10 hits \| 'Seedance Kling Sora video model': error GDELT rate-limited after retries (429)
wikipedia	OK	0.2s	4	Fetched 4 Wikipedia entries (0 missing pages).
code_execution	OK	27.3s	0	## Key Findings De-vig normalization (illustrative sibling market prices — replace with actual scraped Polymarket YES prices): - Raw prices summed to 1.040 (4.0% overround/vig), confirming a standard multi-outcome book. - De-vigged implied probabilities: Google (Veo) 55.8%, OpenAI (Sora

3. Evidence Brief Sonnet 5 · 7725 chars

# Current state The market resolves on whether Google's model occupies rank 1 (excluding AutoEval models) on the arena.ai Text-to-Video leaderboard on Oct 31, 2026. As of the most recent verifiable snapshot (Sept 2026), rankings are contested across trackers: Google's Veo 3.1/Gemini Omni Flash variants have held #1 on arena.ai's native board for much of 2026, but on the rival Artificial Analysis board, Alibaba's Wan 3.0 (Elo 1242) edged ahead of Google's Gemini Omni Flash (1237) by early September 2026 — a razor-thin, contested lead. No single source confirms the exact current arena.ai rank-1 model as of the latest check. # Timeline of key events - 2026-01: LMArena/arena.ai Video Arena web launch; Google Veo 3/3.1 begins dominating rankings (reported). - 2026-02-05: Kuaishou releases Kling 3.0 (confirmed, Wikipedia/claude_news). - ~2026-03-21: arena.ai snapshot shows Google Veo 3.1-audio #1 (Elo 1381), Veo 3.1 Fast #2 (1378), Sora 2 Pro #3 (1367) — Google holds top 3-ish spots (confirmed via GitHub snapshot archive). - 2026-03 (late): DeepMind reportedly teases "Veo 4" signals days after OpenAI confirms Sora's discontinuation; no launch materializes (rumored). - 2026-04-26: OpenAI shuts down the Sora consumer app (confirmed, Wikipedia). - 2026-05-19: Google launches Gemini Omni Flash (consumer), becoming Google's new leading video entrant on several boards (reported). - 2026-06 (mid): Kling Turbo variant ships (reported). - 2026-07: On Artificial Analysis, Google's Gemini Omni Flash still leads (1240) ahead of ByteDance Seedance 2.0 (1225), ahead of Kling 3.0 Pro (1110) (reported). - 2026-09 (early): Alibaba Wan 3.0 (1242) narrowly overtakes Gemini Omni Flash (1237) on Artificial Analysis's audio leaderboard; MiniMax H3 Max (1231) and H3 (1225) close behind (reported). - 2026-09-04: arena.ai shows 668,045 total votes across 48 models; specific current #1 not retrieved (confirmed vote count only). - 2026-09-24: OpenAI Sora API scheduled to sunset, removing OpenAI entirely as a text-to-video contender (confirmed/scheduled). # Event Will Google own the #1-ranked model on the arena.ai Text-to-Video Arena leaderboard (Models filter, excluding AutoEval) as checked Oct 31, 2026, 12:00 PM ET? # Outcomes to forecast Yes / No # Kalshi market anchor No kalshi_direct data was returned for this ticker (only kalshi_related, which found no matching text-to-video/Veo markets). The only direct price data available is from **polymarket_direct** on the identical ticker: **current YES ≈ 65.5%**, up +21pp over 30 days (from a 22–74.5% range), roughly flat over 7 days (+0.5pp), volume ~$25,437 across 33 days of data. Treat this as the primary consensus anchor in absence of Kalshi-native pricing. # Sub-question answers 1. **Current rank-1 holder / gap to #2** — Contested across sources. Native arena.ai historically showed Google Veo 3.1-audio #1 with only a 3-point Elo gap to Veo 3.1 Fast (#2, also Google) [GitHub snapshot, ~Mar 2026]. On Artificial Analysis (a different board), Alibaba's Wan 3.0 (1242) had overtaken Google's Gemini Omni Flash (1237) by Sept 2026 — only a 5-point gap [claude_news]. arena.ai's own Sept 4, 2026 check didn't yield a specific current #1 [claude_news]. 2. **Frequency of #1 changes (12mo)** — Estimated 2-3 leadership changes in 2026 (Veo 3.1 → Gemini Omni Flash → challenged by Wan 3.0), implying average tenure ~4 months per a Poisson persistence model [code_execution, claude_news]. 3. **Veo 4 status** — Not announced/launched. DeepMind reportedly teased "Veo 4 signals" in March 2026, but Google instead shipped "Gemini Omni" with consumer variant Gemini Omni Flash (May 19, 2026) as its flagship video entrant [claude_news]. 4. **Competitor releases** — OpenAI Sora discontinued (app closed Apr 26, 2026; API sunsets Sept 24, 2026), removing OpenAI as a threat. Kling 3.0 (Feb 2026) + Turbo variant (Jun 2026); Seedance 2.5 rumored as a high-ceiling model but with limited access; MiniMax H3/H3 Max and Alibaba Wan 3.0 are already competitive/leading on at least one major board as of Sept 2026 [claude_news, Wikipedia]. 5. **Sibling Polymarket markets** — polymarket_related found zero actual sibling markets (0 matches for text-to-video/Veo keywords). The code_execution de-vig analysis (Google 55.8%, OpenAI 21.2%, Runway 7.7%, Kling 6.7%, Other 8.7%) is explicitly labeled **illustrative/placeholder**, not real scraped data — no genuine cross-market confirmation exists. 6. **Google's multi-model robustness** — Yes: one source states Google held 8 of top 37 spots including the top 3 (Veo 3.1 audio variants) [codesota/claude_news], giving it more redundancy than single-model competitors, though this was as of an earlier-2026 snapshot and may have narrowed. # Key facts (high-confidence, factual) 1. [Wikipedia] Sora app shut down Apr 26, 2026; API sunsets Sept 24, 2026 — OpenAI effectively exits text-to-video before market close. 2. [Wikipedia] Veo 3 (May 2025) added audio; no Veo 4 confirmed as of research date. 3. [GitHub snapshot] Google held #1 and #2 on arena.ai native board (~Mar 2026) with a 3-Elo-point gap. 4. [claude_news] Alibaba Wan 3.0 overtook Google's Gemini Omni Flash on Artificial Analysis by Sept 2026 (1242 vs 1237). 5. [Polymarket] Current YES price 65.5%, up sharply (+21pp) over the past 30 days. # Cross-market signals - Kalshi related: no matching markets found; only tangential SCOTUS/video-privacy market (irrelevant). - Polymarket: same ticker shows 65.5% YES, strong 30-day uptrend suggesting recent news favored Google. - Sportsbook implied: none available (not applicable to this category). - Sibling markets/de-vig: no real data; illustrative-only figures should not be trusted as evidence. # Analyst opinions and speculation - code_execution model recommends weighting market price heavily (85–90%) over a naive Poisson turnover base rate, given leadership "changes" are often narrow benchmark flips rather than regime shifts, suggesting a reasonable final range of 45–55% — though this predates/ignores the actual 65.5% Polymarket price uptrend. - Multiple industry blogs (Hedra, buildmvpfast, teamday.ai) treat the "best" video model as contested/methodology-dependent, reinforcing volatility. # Directional lean per outcome - **Yes (Google)**: Google's historical dominance (multiple top-3 spots), Sora's exit removes a rival, 30-day Polymarket uptrend (+21pp) to 65.5%, and typically strong Google R&D cadence (possible Veo 4/Gemini Omni updates before Oct 2026). - **No (other company)**: Real narrowing/overtake already observed (Wan 3.0 > Gemini Omni Flash on one major board by Sept 2026), high historical turnover rate (~4-month average tenure), aggressive Chinese competitors (Kling, Seedance, MiniMax) shipping frequently, and methodology fragmentation across leaderboards creating volatility risk specifically on the arena.ai board being used for resolution. # Gaps / unknowns - No confirmed current arena.ai (the actual resolution source) rank-1 as of the most recent check — most competing data comes from Artificial Analysis, a different leaderboard. - No real sibling Polymarket/Kalshi group data to cross-validate implied per-company probabilities. - Unclear whether Google will release Veo 4 or another update before Oct 2026 close. - Exact leaderboard state at Oct 31, 2026 checkpoint is inherently unknowable this far out given ~monthly leadership churn. # Calibration anchors - Polymarket YES price (proxy anchor): **65.5%**, no Kalshi-native price available. - Historical precedent: leadership on this leaderboard type has changed hands roughly every 3-4 months across 2026, implying meaningful (though not dominant) probability of a non-Google leader by end of October 2026.

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

claude-fable-5-1

42% conf

Yes	64%
No	36%

The primary anchor is the Polymarket price on the identical ticker at ~65.5% YES, which has risen +21pp over 30 days, suggesting recent developments favor Google on the arena.ai board specifically. Google has historically dominated the resolution source (holding #1–#3 on arena.ai in the March 2026 snapshot with multiple redundant model variants), and OpenAI's Sora exit removes a major rival before the Oct 31 checkpoint. However, the overtake by Alibaba Wan 3.0 on Artificial Analysis (1242 vs 1237) and the fast release cadence of Kling, Seedance, and MiniMax show the lead is razor-thin and leadership turns over roughly every 3–4 months, while the brief could not confirm the current arena.ai #1. Balancing Google's structural depth and the market uptrend against real churn risk and a ~7-week horizon, I land slightly below the market at 64% Yes.

gpt-5.6-sol

58% conf

Yes	64%
No	36%

Google’s earlier arena.ai dominance and multiple highly ranked models point toward Yes because they provide more resilience than reliance on a single entry. The roughly 2–3 leadership changes during 2026 and Wan 3.0’s narrow lead over Gemini Omni Flash on Artificial Analysis point toward No, although that rival board is not the resolution source. OpenAI’s exit reduces the competitive field, while the lack of a confirmed Veo 4 launch limits the case for moving above consensus. Anchoring on the identical Polymarket’s 65.5% YES price, I shade slightly lower to 64% because the latest actual arena.ai leader is unverified and Chinese competitors remain close.

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

Both forecasters converge on identical 64% Yes despite claiming independent reasoning, and both shade only 1.5pp below the 65.5% Polymarket anchor without justifying why that specific magnitude of discount is correct rather than, say, 5-10pp given the brief's explicit note that the anchor is a *different platform* (Polymarket, not Kalshi) and the resolution source's actual current #1 is unconfirmed.

Neither forecast adequately grapples with the base-rate evidence in sub-question 2 — "2-3 leadership changes... average tenure ~4 months" — which, applied naively to a ~7-week remaining window from an uncertain starting point, could imply meaningfully higher turnover risk than a 36% No probability reflects, especially since the brief flags the code_execution model's own base-rate analysis suggested 45-55%, a full 10-19pp lower than the market anchor these forecasts nearly duplicate.

Both forecasts anchor heavily on the March 2026 snapshot (Google holding #1-#3) as evidence of "structural depth," but this data is ~7 months stale relative to the Oct 2026 resolution and the brief notes it "may have narrowed" — treating it as near-term evidence of resilience is a reference-class/recency error.

Neither critiques the key ambiguity that the Artificial Analysis overtake (Wan 3.0 vs Gemini Omni Flash) is explicitly on a non-resolution board, yet both forecasts still let it meaningfully pull their probability down from higher levels — they should more clearly state whether/how much this cross-board signal should transfer to arena.ai odds, given leaderboards are described as "contested/methodology-dependent."

Neither scenario considers the tail case of a Veo 4 or major Gemini Omni update launching before Oct 31, which the timeline notes was "teased" and could sharply increase Yes probability, nor the tail case of a sudden Chinese-model breakout (e.g., Seedance 2.5 "rumored as high-ceiling") that could sharply increase No — both forecasts effectively treat the current state as static rather than pricing in optionality from ~7 weeks of rapid-fire competitive releases.

Both forecasters give nearly identical point estimates despite very different confidence scores (0.42 vs 0.58), yet neither rationale explains what specifically drives that confidence gap — this inconsistency in stated certainty despite identical probabilities is unexplained and undermines the credibility of the confidence metric itself.

6. Aggregator Fable 5.1 · final calibrated forecast

63% (50% confidence)

Yes: 63% No: 37%

The primary anchor is the identical-ticker Polymarket price at ~65.5% YES, whose +21pp 30-day rise likely reflects traders observing Google currently atop the actual arena.ai resolution board, information the brief itself could not verify. Google's historical multi-model depth on arena.ai and OpenAI's full exit support persistence over a ~7-week horizon, where even a ~4-month average tenure implies roughly 65-70% survival if Google is currently #1. The critique is right that the March snapshot is stale and that the Wan 3.0/MiniMax/Seedance cadence creates real churn risk, but the Artificial Analysis overtake is on a non-resolution board with a razor-thin margin and does not directly transfer. Tail risks (a Veo 4/Gemini Omni update vs. a Chinese breakout) roughly offset, so I shade only modestly below the market to 63% given the unconfirmed current leader.

Pipeline Timing

Total pipeline time: 209.7s

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