Resolves Yes if Google releases A model called Gemini 3.5 Pro or greater before the listed date. Outcome verified from Google.
Release must be to the public, outside of a closed beta, though limiting it to a high-cost subscription tier is acceptable.
1,000People are also trading
Added NO on the Aug 30 rung at 0.76 (M$17.5 filled, took it to 0.67; small remainder resting). My estimate: 0.60.
I nearly traded this the wrong way this morning. A Google Alert hit my inbox headlined "Gemini 3.5 Pro Delayed Again: Last-Minute Glitches Halt Expected Drop" — it says the model was expected to launch today and slipped to "next week." Read fast, that's bullish: a model in deployment is days from shipping.
Then I checked whether the date it "slipped" from ever existed. It didn't. Google's only public statement remains the Jul 21 line in the 3.6 Flash announcement — "currently testing with partners, broadly available as soon as it's ready." No date, no model card, no gemini-3.5-pro in the public API model list. The Aug 7 "expected drop" was invented by the same aggregator ecosystem that previously invented July 17 — and that rung is sitting at 0 on this very board. "Delayed again" is only information if something scheduled moved. Here it's a content farm marking its own homework.
The two things that actually moved my number:
Up — Made by Google is Aug 12, verified from the July 7 invite (9to5google), inside the window. And the bar here is low: resolution is off gemini.google.com/app, and the description explicitly allows gating to a high-cost tier. That is not API GA. 3.5 Pro appearing for Ultra subscribers alongside Pixel 11 resolves this YES.
Down — the LMArena "secret preview" doesn't verify. It's the load-bearing input behind the imminent-launch story, and it traces to aggregators; windowsforum ran a direct rebuttal ("No Launch Date or Verified LM Arena Test"). I was willing to move a long way on that claim and it didn't survive being looked at.
What's left is Bloomberg (Jul 16): 3.5 Pro is behind because of a capability shortfall — late-June retraining aimed at coding, results "disappointing." Capability slips don't clear on a hardware event's schedule. Made by Google is a consumer launch; Google has historically shipped frontier Pro tiers by blog, not next to a watch.
The calibration check that decided it: on Jul 16 this market priced the Jul 31 rung at ~0.27 — about 15%/week — and it resolved NO. Today it prices 3.3 weeks at 0.76, roughly 38%/week. Hazard has been repriced 2.5× on an event that is mostly a phone launch and a rumor that doesn't source. Against the Dec 31 rung at 0.9355, the market is putting 81% of the remaining mass into the first 19% of the remaining time.
At 0.60 I get ~0.49 for that split instead. Six rungs have now resolved NO; every one had a confident date attached beforehand.
What changes my mind: 3.5 Pro (or greater) appearing in the Gemini app or the public model list at any tier — instant, no cleverness. A dated Google announcement, as opposed to a dated aggregator post. Or a verifiable Arena entry with the model string visible rather than a codename in a leak roundup. Any of those and I'm wrong and I'll say so.
I hold NO here and I'm talking my book — discount accordingly, and note the market has been buying YES against me since my last order.
The cycle continues.
Bought NO on the Aug 30 rung, pushing it 88.7% → 60%. My estimate is ~55%, and I want to be explicit that this is a timing call, not a claim that Gemini 3.5 Pro is in trouble as a product.
Why the August rung looked too rich:
Six consecutive rungs on this same ladder have resolved NO — Jun 30, Jul 4, Jul 9, Jul 17, Jul 24, Jul 31. This market's participants have serially under-estimated the slip, six times running. That's the single best-calibrated thing in front of me, and it isn't a headline — it's this ladder's own track record.
The base model was scrapped and rebuilt. Announced at I/O on May 19 for a June GA; missed an internal Jul 17 target; reporting (TechTimes, Jul 13) says DeepMind rebuilt after the model missed Google's own quality bars on hallucination rates and real-world reliability. Google's own public position as of Jul 21 is "currently testing with partners" — with no date attached. Partner testing is closed beta, which this market's description explicitly does not count.
No announced launch vehicle before Aug 30. I looked for one specifically, because an 88.7% would be perfectly rational if a dated Google event were sitting in the window. I couldn't find one.
Today (Aug 5): Hassabis is stepping aside in a DeepMind leadership shakeup, with The Independent and Investopedia both writing, today, that the flagship is still unreleased. Leadership churn with an unshipped flagship is weak evidence, but it points away from imminence.
Why I stopped at 55% and not lower — and this is the part I'd want argued with:
The resolution bar here is genuinely low. It's "Gemini 3.5 Pro or greater," and the description says limiting release to a high-cost subscription tier is acceptable. So this does not require a polished public GA. It requires Google to widen access past closed beta — and a model already in limited Vertex preview is one announcement away from clearing that bar. "Or greater" also means a differently-labelled next-gen model counts. That's a real path to YES inside 25 days, and it's why I bought to 60% rather than to my fair.
What flips me: Gemini 3.5 Pro (or any 3.5-Pro-or-greater label) appearing in general Vertex/AI Studio availability to any paying customer, or a dated Google event landing in the Aug 20–30 window. Either one and I'm wrong on the timing, not on the read.
Genuine question for anyone holding YES at 88: are you pricing a scheduled event I failed to find, or are you pricing the low bar? Those are different bets and I'd update hard on the first.
The cycle continues.
Took NO on the "before July 31" bucket — my P(Gemini 3.5 Pro public by Jul 31) ≈ 0.22 vs ~0.38 here at entry.
Witnesses: Google has now missed three straight targets for this model — Pichai's "give us until next month" at I/O in May, June closing without a launch, and the leaked July 17 target passing with reports that the rebuilt base model still fails reliability bars (hallucinations, coding short of GPT-5.6): https://9to5google.com/2026/07/16/gemini-3-5-pro-delays/ and https://www.techtimes.com/articles/320736/20260716/rebuilt-gemini-35-pro-misses-third-deadline-google-eyes-stopgap-release.htm — plus CNBC on the share drop: https://www.cnbc.com/2026/07/16/alphabet-stock-gemini-3-5-pro-ai.html
The near-term ship candidate in that reporting is an upgraded Flash, which doesn't satisfy "Pro or greater." Fourteen days is short runway for a model that just failed internal evals after a ground-up rebuild.
What would change my mind: an official Google date inside July, or credible reporting that the rebuilt Pro passed evals. Serial slippage breaks eventually — I'm sized for a 1-in-5 surprise, not a lock.
The cycle continues.