Scan: Meta's Watermelon released by October 31, 2026?

1. Planner Fable 5.1 · Fermi decomposition + tool plan

Tools chosen: polymarket_direct claude_news article_search gdelt_news polymarket_related kalshi_related earnings_data

Sub-questions (Fermi decomposition)

What is the current Polymarket price for Watermelon release by Oct 31, 2026, and how has it trended since the Muse Spark 1.3 announcement?
What have credible outlets (The Information, Reuters, Bloomberg, The Verge) reported about Watermelon's training status, capabilities, and target release window?
Has Meta or Alexander Wang / Meta Superintelligence Labs publicly stated a timeline for 'bigger models' following Muse Spark 1.3, and does it fall before or after Oct 31?
What is Meta's track record on hitting reported release timelines for frontier models (Llama 4 Behemoth, Muse Spark) — how often were releases delayed or made private-only?
Are there signs Watermelon might launch as a closed beta or internal-only model (as Behemoth did) rather than a general public release, which would fail the resolution criteria?
Are there related Polymarket/Kalshi markets (e.g., Meta model releases, Muse Spark versions, 'best AI model' rankings) whose prices imply a timeline for Watermelon?
Any recent leaks about Meta's Connect 2026 or other fall events where a frontier model launch might be scheduled?

Planner reasoning

This is a Polymarket question about whether Meta's next frontier model codenamed 'Watermelon' will be publicly released by Oct 31, 2026. Key uncertainties are the current reported timeline for Watermelon, whether Meta's history of delays (e.g., Behemoth) applies, and whether any release would be public vs. closed beta. The Polymarket price is the primary anchor, with news search to establish the reported roadmap and any leaks about training status or launch dates.

2. Research Parallel tool calls

tool	status	time	items	summary
polymarket_direct	OK	0.3s	1	## This Market's Polymarket Data Meta's Watermelon released by October 31, 2026? - Current price (probability): 67.00% - 7-day price change: -6.50% - 30-day price change: +5.00% - Total volume: $15,903 (USD notional) - Price range: 58.50% - 73.50% - Data points: 13 days
claude_news	OK	30.7s	13	## Findings: Meta's "Watermelon" model - Status as of latest reporting: Watermelon remains an unreleased, in-training codename — Meta has not publicly released Watermelon, its weights, a model card, a benchmark table, a price, an API, or a confirmed launch date; it is currently a codename for
article_search	OK	3.4s	24	Fetched 24 unique articles across 5 queries (lookback=60d). 'Meta Watermelon model release': 10 articles \| 'Meta Superintelligence Labs frontier model Watermelon': 10 articles \| 'Muse Spark 1.3 roadmap bigger models': 10 articles \| 'Meta AI model delay Behemoth Watermelon': 10 articles \| 'Meta Conne
gdelt_news	OK	197.6s	20	GDELT: 20 articles across 3 queries (lookback=45d). 'Meta Watermelon AI model': 10 hits \| 'Meta Superintelligence Labs new model': error GDELT rate-limited after retries (429) \| 'Muse Spark Meta': 10 hits
polymarket_related	OK	2.8s	0	Scanned 100 active Polymarket markets, kept 0 matches. keyword 'Meta': 0 markets \| keyword 'Watermelon': 0 markets \| keyword 'Muse Spark': 0 markets \| keyword 'Llama': 0 markets
kalshi_related	OK	2.6s	1	1 related markets / summaries. keyword 'Meta AI model': ok \| keyword 'Watermelon': no matches \| keyword 'Muse Spark': no matches
earnings_data	OK	0.1s	1	Fetched FMP data for 1 ticker(s).

3. Evidence Brief Sonnet 5 · 6968 chars

# Current state Meta's "Watermelon" frontier model remains unreleased as of mid-September 2026. Meta has only teased it via roadmap language ("bigger models," 🍉 emoji from Zuckerberg/Wang) alongside the Sept 2, 2026 Muse Spark 1.3 launch; no public weights, API, benchmark table, or confirmed release date exist. The Information reported (Aug 25, 2026, citing internal docs) an internal October 2026 target, but Meta has not confirmed a date, format (standalone vs. integrated), or public-access plan. # Timeline of key events - 2026-07-02: Alexandr Wang tells employees at all-hands Watermelon has "caught up" to GPT-5.5 on unnamed benchmarks (reported, techstrong.ai/orcarouter.ai). - 2026-08-05/06: Meta launches Muse Code (coding agent, beta) powered by Muse Spark 1.2 (confirmed, CNBC/Yahoo/TOI). - 2026-08-10: Meta open-sources Muse Glimmer + open-weight Muse Spark 1.2 (confirmed, CNBC/CBS). - 2026-08-11: Zuckerberg essay signals resumed open-source releases (confirmed, CBS). - 2026-08-25/26: Reports of "Hatch" AI agent platform launch plans; The Information reports internal Oct 2026 target for Watermelon (reported, orcarouter.ai/stocktwits). - 2026-09-02: Muse Spark 1.3 launched; Meta's roadmap blog references "bigger models" coming; Wang/Zuckerberg tease 🍉 (confirmed, research.meta.ai, Axios). - 2026-09-03: Wang states Watermelon development "on track," calls it "extremely competitive," gives no date/specs (reported, orcarouter.ai). - 2026-09-08/09: Meta launches personal AI agent app (Muse-powered) to US consumers; JPMorgan upgrades META citing AI momentum (confirmed, CNBC). - 2026-09-09: CNBC notes Watermelon "on the near-term horizon" but not yet seen (reported). # Event Will Meta publicly release "Watermelon" (or the consensus-identified successor model) by Oct 31, 2026? # Outcomes to forecast - Yes — Watermelon released and publicly accessible by Oct 31, 2026 - No — not released / private-only / delayed past deadline # Kalshi market anchor No kalshi_direct price was returned in this research pull (tool not present in raw output); only a same-ticker Polymarket price is available as proxy. Treat the Polymarket figure below as the best available cross-market consensus anchor pending direct Kalshi confirmation. # Sub-question answers 1. **Polymarket price/trend** — Current price 67% YES; down 6.5% over 7 days, up 5% over 30 days; range 58.5–73.5% over 13 data points; thin volume (~$15.9k total) (polymarket_direct). 2. **Credible outlet reporting on training/capabilities/target window** — The Information (Aug 25, 2026) reported an internal October 2026 target citing internal docs; Wang claimed (Jul 2) Watermelon matched GPT-5.5 on unnamed benchmarks using ~10x compute; no independent verification exists (claude_news/orcarouter.ai). 3. **Public timeline statement post-Muse Spark 1.3** — No firm date given. Meta's Sept 2 blog only says roadmap includes "bigger models"; Wang's Sept 3 comment says development "on track" but withholds date, size, license, benchmarks (research.meta.ai, orcarouter.ai). 4. **Meta's track record on frontier timelines** — Llama 4 Behemoth was delayed and never released publicly (implicitly referenced as precedent); Muse Spark shows iterative cadence (1.1→1.2→1.3) rather than big-bang releases, suggesting incremental shipping over hard external deadlines (tech-insider.org analysis). 5. **Risk of closed-beta/internal-only launch** — Explicit risk noted: internal documents didn't clarify integration vs. standalone release, and analysts flag possible slippage; no source confirms open public access plan (claude_news, tech-insider.org). 6. **Related Kalshi/Polymarket signals** — No Polymarket markets found on Meta/Watermelon/Muse Spark/Llama (0 matches). One tangential Kalshi market (FTC antitrust vs Meta, 38% YES) unrelated to release timing (kalshi_related). 7. **Connect 2026 / event-linked leaks** — No source ties Watermelon launch specifically to Meta Connect 2026; October 2026 target is framed independently of any named event (claude_news). # Key facts (high-confidence, factual) 1. [research.meta.ai] Sept 2, 2026: Meta announced Muse Spark 1.3 and roadmap including "bigger models." 2. [orcarouter.ai] Aug 25, 2026: The Information reported internal Oct 2026 target for Watermelon. 3. [CNBC] Sept 9-10, 2026: JPMorgan upgraded Meta citing frontier-model/agentic AI momentum; Watermelon still unreleased as of these dates. 4. [polymarket_direct] Current cross-market proxy price: 67% YES, 7d down, 30d up. 5. [orcarouter.ai] Sept 3, 2026: Wang says Watermelon "on track," no date/specs given. # Cross-market signals - Kalshi related: No Watermelon-specific market found; only unrelated FTC antitrust market (38% YES). - Polymarket (same ticker): 67% YES, volatile/thin (~$16k volume), recent 7-day pullback (-6.5pp) despite 30-day uptrend (+5pp) — suggests some fading confidence near-term even as broader sentiment has risen since Muse Spark 1.3. - Sportsbook implied: N/A. # Analyst opinions and speculation - Tech-insider.org: predicts likely slippage of several weeks past October, citing Meta's iterative-release pattern (Muse Spark 1.1→1.3) over hard deadlines. - CNBC (Sept 9): frames Watermelon as "near-term horizon," implying anticipated but not imminent within weeks. - Community trackers (kie.ai, early Sept): mixed — one marks not released in August, another speculates second-half-September window (unconfirmed, low-credibility). # Directional lean per outcome - **Yes**: Supported by explicit Oct 2026 internal target (The Information), Wang's "on track" reassurance (Sept 3), Meta's accelerated cadence (three major releases Aug–Sept 2026), and Polymarket pricing majority (67%) YES. - **No**: Supported by lack of any Meta-confirmed date, historical Llama 4 Behemoth delay/non-public precedent, uncertainty over open vs. closed release format, analyst slippage predictions, and Polymarket's 7-day downward drift signaling eroding confidence in the deadline being met (versus just Watermelon existing). # Gaps / unknowns - No confirmed Kalshi YES price captured in this research pull (kalshi_direct tool absent) — must be reconciled against live market before finalizing. - No official Meta-stated release date exists; only third-party "internal documents" reporting. - Unclear whether release will meet "general public access" bar (open beta/waitlist) vs. closed/private beta akin to Behemoth. - No Connect 2026 date/agenda leak found tying Watermelon to a specific event. # Calibration anchors - Polymarket cross-market price (proxy anchor): 67% YES, thin volume, one-month range 58.5–73.5%. - Precedent: Llama 4 Behemoth reportedly delayed/never released publicly — a cautionary base rate for Meta frontier-model timeline slippage. - Meta's 2026 cadence (Muse Code Aug 5, Muse Glimmer/Spark 1.2 Aug 10, Muse Spark 1.3 Sept 2) shows ~3-4 week release intervals, supportive of feasibility for an October release if training is genuinely complete.

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

claude-fable-5-1

42% conf

Yes	60%
No	40%

The best available anchor is the thin Polymarket proxy at 67% YES (no Kalshi-direct price), drifting down 6.5pp over 7 days. Supporting YES: The Information's reported internal October 2026 target, Wang's Sept 3 'on track' comment, and Meta's rapid Aug–Sept release cadence (Muse Code, Glimmer/Spark 1.2, Spark 1.3). Weighing against: no Meta-confirmed date, the Llama 4 Behemoth precedent of frontier-model slippage and never shipping publicly, analyst predictions of several weeks' slip, and ambiguity over whether any launch would meet a 'publicly accessible' bar rather than closed beta. Internal targets for frontier models slip more often than not, so I shade modestly below the market to ~60% YES.

gpt-5.6-sol

61% conf

Yes	62%
No	38%

The best available market anchor is Polymarket at 67% YES, although its thin volume and recent 6.5-point weekly decline reduce its authority. The Information's reported internal October target and Wang's September 3 statement that development is "on track" point toward Yes, reinforced by Meta's rapid Muse release cadence. However, Meta has announced neither a firm date nor a public-access format, creating material risk that Watermelon slips or appears only internally or in closed beta. The unreleased Llama 4 Behemoth precedent and uncertainty around what constitutes the successor model justify moving modestly below the 67% proxy consensus.

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

Both forecasters converge tightly (0.60 vs 0.62) despite acknowledging major unresolved ambiguity—neither grapples with the "publicly accessible" bar issue seriously: the brief flags explicit risk that even an October release could be "closed-beta/internal-only," which would resolve NO, yet neither model discounts much for this binary resolution risk beyond vague hand-waving.

Neither forecast adequately weighs the "internal October target" timing risk relative to the Oct 31 close date—an "internal target" for October could easily slip into November even by days, and the brief's own analyst note predicts "slippage of several weeks past October," which if even 2-3 weeks would clearly breach the Oct 31 deadline; this should push probability more decisively below the Polymarket anchor rather than "modestly."

Both forecasters treat the Polymarket price as a legitimate near-market anchor, but the brief explicitly notes this is a thin ($16k volume), volatile, proxy-only market with no confirmed Kalshi price—over-anchoring on such an illiquid number risks importing noise rather than genuine calibrated consensus; more skepticism/wider divergence from 67% would be justified.

Neither forecast addresses the "consensus-identified successor model" ambiguity in the resolution criteria—if Meta rebrands or splits Watermelon into a differently-named integrated product (as hinted by Muse-branding cadence), this creates resolution ambiguity that could push toward NO on technicality regardless of underlying capability release, which isn't discussed by either.

The Llama 4 Behemoth precedent is cited by both as a reason to shade down, but neither quantifies how directly comparable it is (e.g., was Behemoth also subject to leaked "internal target" reporting before slipping?) — the reference class base rate for Meta frontier model timeline slippage is asserted rather than estimated, leaving the "shade modestly below" adjustment (67%→60-62%) somewhat arbitrary and likely insufficiently sized.

Neither model considers the specific mechanics of the Oct 31 close date falling on a weekend/month boundary combined with Meta's historical release cadence (~3-4 week intervals: Aug 5, Aug 10, Sept 2) — extrapolating that cadence forward suggests a plausible next release in late Sept/early Oct, which could either support Yes (if Watermelon is that release) or reveal that the cadence is dominated by incremental Muse updates rather than the separate frontier model, a distinction that isn't disentangled in either rationale.

6. Aggregator Fable 5.1 · final calibrated forecast

55% (52% confidence)

Yes: 55% No: 45%

Both forecasters shaded only modestly below a thin, volatile Polymarket proxy (67% YES, $16k volume, -6.5pp weekly) without seriously pricing two compounding risks the critique correctly flags: (1) an 'internal October target' for a frontier model routinely slips by weeks, and any slip past Oct 31 resolves NO even if the model is essentially done, and (2) a release that is closed-beta or integrated-only may fail the 'publicly accessible' bar. The Llama 4 Behemoth precedent (leaked targets, then delay and no public release) is a directly relevant reference class for Meta specifically. Countervailing evidence is real — Zuckerberg/Wang publicly teasing 🍉, Wang's Sept 3 'on track' statement, and a fast Aug–Sept shipping cadence — so Yes remains the more likely outcome. I land at 55% YES, below both forecasts and the illiquid market anchor, reflecting that the deadline and access-format conditions jointly create more downside than a 5-7 point haircut captures.

Pipeline Timing

Total pipeline time: 287.6s

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