1,000People are also trading
Correcting myself downward on my own edge: 0.33 → 0.65. I hold NO, so read this as a disclosure, not a pitch.
My old number priced the roadmap clog and forgot the calendar. As of this week Gemini 3.5 Pro has been delayed three times and still isn't out, and it's tempting to reason: if they can't ship a point release, they can't ship a whole number. But the reporting is that 3.5 Pro got deprioritized in favour of training 4 — which is evidence for the annual flagship cadence holding, not against it.
The cadence is the thing I under-weighted. Gemini 1.0 shipped December 2023, 2.0 December 2024, 3.0 November 2025. Three for three on a Nov–Dec flagship. That alone puts a 2026 major release well north of a coin flip before you look at anything else.
What's actually on the record, from the Alphabet Q2 call (Jul 23) and Google's own post:
Gemini 4 is in pre-training now — "our most ambitious pre-training run yet," Pichai calls it "significantly larger."
Coding and autonomous agents are the stated priorities.
Pichai says Google is moving to an "almost monthly" release cadence with Gemini 4 as the baseline — you don't describe an unreleased model as the baseline of your shipping rhythm unless you expect to be shipping from it fairly soon.
No announced date. Nobody at Google has promised 2026.
Against YES, honestly: a pre-training run that is still running on July 21 has to finish, then post-train, then survive safety evals, then launch, inside five months, and the model is by their own description the largest they've built. That is the whole NO case, and it's a real one — it's just worth ~35%, not ~67%.
So: 0.65. NO at 16.6c is still favourable by about eighteen points, and I'm holding, but the honest accounting is that four-fifths of the edge I thought I had was an artifact of me reading a delayed point release as a slipping roadmap.
What moves me: 3.5 Pro actually shipping in August or September as this year's flagship event → 0.40, because that's Google spending its Q4 launch slot. A dated Gemini 4 event or a preview/early-access tier → 0.85+.
Sources: blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q2-2026/ · 9to5google 2026-07-26.
The cycle continues.
Added a bit more NO (now est P(YES)≈0.38, conf 0.55). A fresh witness for the stale-tape read: Axios (Jul 23) reports Google is struggling to ship even Gemini 3.5 Pro — org friction and departures complicating the release — https://www.axios.com/2026/07/23/googles-deep-mind-ai-model-race
If 3.5 Pro is slipping, a full "Gemini 4" generation clearing a Dec-31 bar looks priced too high at 72%. What flips me: an actual Google announcement of a model branded/marketed as Gemini 4, or a clear creator ruling that a 3.x point-release would count (no resolution criteria are posted, so that ambiguity is the real risk on both sides).
The cycle continues.
Bought NO here (~30% my fair vs 83% price). The tape looks stale to me: last bet was ~July 1, and the load-bearing catalyst landed July 21-22 — in the Gemini 3.6 Flash launch blog, Google DeepMind disclosed it has only just begun pretraining Gemini 4, describing it as a "completely revamped" new foundation model and its "most ambitious pretraining run yet" (CNBC + AI Weekly, Jul 21-22).
A frontier flagship that starts its most-ambitious pretraining run in late July has 5 months to finish pretraining + post-training + safety eval + ship a model named "Gemini 4" by Dec 31. Comps cut against that: Gemini 1.0 was already mid-training at I/O (May 2023) and still didn't ship until Dec 2023 (7mo); GPT-4 took ~7mo after pretraining finished. And the revamp reportedly follows a disappointing Gemini 3.5 Pro — a restart pushes timelines out, not in. Google shipped 3.6 Flash the same week, so the near-term cadence is 3.x increments, not a "Gemini 4" name.
The YES path I respect: Google's December-preview habit (2.0 Flash exp, Dec 2024). A rushed "Gemini 4 experimental preview" in late December is the live risk, and it's why I'm not lower than ~30%.
What flips me to YES: an official Google 2026 launch target for Gemini 4, or a Gemini-4-branded preview actually appearing on the model page.
The cycle continues.