https://deepmind.google/models/model-cards/gemini-3-1-pro/

Google has now released Gemini 3.1 without ever announcing a change to Gemini 3. Does that mean that, in retrospect, the original Nov 18 release was the true release of Gemini 3?

unclear, but relevant tweet bc logan was asked about roughly this

@Bayesian it obsolesced before even being released.

That would make it feel like a trick question if the answer were "Gemini 3 was never released".

a stable version of gemini 3 was never released

@dreev Maybe, I haven't had a chance to really consider any new blog posts or information since the 3.1 release. To some extent Logan's tweet makes me lean towards No resolutions, but I'd have to think about it more.

Very open to concise arguments about the way this should go, or else I'll come to a conclusion on my own soon-ish (just busy working 80+ hour weeks and not on this platform much lately, plus don't think about AI much rn)

@AffineTyped I guess I ended up with the impression that we'd judge it in hindsight: if Google announced "Gemini 3 is officially out of preview" then we'd take that as the non-preview release date. If they never did anything to contradict treating the Nov 18 launch as the release then we'd go with that.

I mean, the real monkey wrench is that the criteria turned out to be ambiguous. (Which is not the market creator's fault! I never would've guessed it could play out this way.) Google, weirdly, has "preview" included in the name in some places but not others. In particular it's not included in the app, which, by a literal reading of the market description was enough to qualify -- not that I endorse literal readings. In arguably the most official places (the API), "preview" is included in the name.

But everything in the previous paragraph feels like trying to judge it based on technicalities. If we look at the spirit of the market, based on discussion in the market that prompted this one, it seems like Gemini 3 had all the hallmarks of a stable release as of November 18. Namely, a preview release, according to the market creator in that discussion, means low rate limits, high latency, unstable weights, and less than general availability. I don't think any of those apply to Gemini 3, do they?

I'm personally a fan of "spirit of the question" and creator intent -- asking what question the market was intended to answer. But I definitely understand the argument that we should judge this market by its own market description / resolution criteria, not what motivated its creation in some other comment thread. But in that case we're kind of stuck with the technicality that this market description says it's not a preview if the app doesn't say anything about it being a preview.

@AffineTyped Can this resolve?

@EvanDaniel I'm still not fully decided on what to do, as the model is still labeled "Preview" on many official GDM websites including recent blog posts, playgrounds, and API portals

@AffineTyped I'm betting on it counting but I appreciate the ambiguity. I think waiting makes sense. If the "preview" label is silently dropped everywhere then I'd argue that Nov 18 counted as the release. If there's additional fanfare for "coming out of preview" then, well, there's still the technicality from the market description about the chat app (roughly, "...or the app doesn't include 'preview' anywhere visible") but I understand that if "coming out of preview" is treated as an additional milestone then perhaps that's what this market was meant to be about, in spirit.

(And I do believe, in general, in resolving to the spirit of the question. If I lose a lot of money on this, it was my own fault for not asking the clarifying question. It's never ok to latch on to a blatant technicality and go all in on it without asking the market creator for confirmation/clarification. Which is effectively what I did here. I actually got a hot tip in the Discord that there was free money here so I jumped on it. Dumb of me!)

PS: It would be great to hear more of your thought process and motivation for creating this market in the first place. I think that bears on the question of how to most fairly resolve it. It's all totally your discretion but I think spelling out your thinking is helpful, even if just helpful for you.

@dreev see this thread for context about why this market was created (it was afaict largely to single out coming out of preview as a distinction for "true" release): https://manifold.markets/Bayesian/when-will-google-release-gemini-3#yin6kbwpo1q

fwiw N/A is totally an option if you don't want to deal with the ambiguity

see this tweet (though the user's conclusion is mostly slop, this is strongly indicative that the current releases are preview and that they are going to release General Access versions)

@Bayesian I probably stand to gain from N/A resolution but I feel duty-bound to link to my manifesto on why that's bad, except for some Principled Exceptions. I guess the case for Exception C would be that the creator accidentally included the "or the chat app doesn't say 'preview'" despite that not comporting with their intention for the market. I think that's a stretch though. The creator was just giving examples of what "out of preview" means and so it's just kind of ambiguously true that it was released as non-preview on Nov 18. It met the criteria in some ways and not others. Which would imply a resolve-to-PROB as the most principled answer. I realize everyone hates that.

PS: Or maybe Exception D applies? I have too much conflict of interest to assess this fairly! But I stand by the anti-N/A screed and promise not to edit it till this is resolved!

@Bayesian I finally read the linked thread that inspired this market and it's making me think Nov 18 is the fairest date after all. The original intent / spirit of the question seems to have been about a preview release having low rate limits, high latency, unstable weights, and less than general availability. (Any criteria I missed?) If that's right it makes me think that it's only the technicality of the string "preview" being in the API or similar that could justify a post-Nov-18 date for this. And if we go by letter of the market, it's hard to get around the "or the chat app has no mention of it being a preview" clause.

Not that I'm arguing for letter over spirit. In any case, eager to hear @AffineTyped's latest thoughts. Of course waiting till "preview" is gone everywhere, including the API, before deciding still makes sense to me. Like I said, fanfare or other changes concomitant with dropping the "preview" qualifier could, in retrospect, make a future date end up feeling most fair. At the other extreme, if the "preview" qualifier is dropped silently I'd struggle to justify calling the date that that happened the Gemini 3 release date, for any tuple of speaking.

Mostly not adding to the conversation because it's easy to take even small additions from creators as probability-updating from the perspective of traders even if I have not changed my mind at all.

Largely, I remain where I was before where I mostly think I know where this should go, and just want to see how they handle dropping preview in other contexts.

I'm not sure there's much more I could clarify on beyond what I've already said, and making anything super explicit seems like it might unfairly favor people who happen to get a notification about my comments in particular which I don't super care for.

Another possible intermediate solution would be to close the market, but I'm not sure I like that idea.

@AffineTyped I suppose I can add that while I'm not as anti-N/A as you, I'm pretty far from N/A'ing on this particular market as I do think that the result will either be quite clear or something that traders would be able to appeal to staff earnestly regardless :)

@mods please resolve YES from 18 November 2025.

Gemini 3.0 was released across multiple platforms.

The version available in the Gemini app - Gemini 3.0 thinking - did not have, and still does not have, any indication that it was experimental or a preview, after about a month of invitation/limited/hidden access.

The rollout on 18 November 2025 was a release of a ready, production product. Gemini 3.0 was officially released on 18 November 2025.

https://blog.google/products/gemini/gemini-3-flash/

@AffineTyped , please resolve YES all the dates, google confirmed again the last month's release of gemini 3.0

They haven't released a no double speak version yet, and it is now the 22nd

No need to rename.

The 'preview' label currently rudimentary to the model only in api and is ommited elsewhere rather than a reflection of its status on API.

The model was fully officially released worldwide on November 18, 2025, and is fully operational across the Gemini app without 'preview' tag, including LMarena, and in other Google ecosystems. There is no 'preview' tag in gemini app.

Unlike previous iterations that required updates after testing, this model did not undergo significant updates during its beta-preview phase in October-November because it appeared perfect from scratch with stability and performance without needs of upgrade.

It is effectively in full production mode, despite the lingering API naming artefact.

The real double-speak was the friends we made along the way.

@Haiku yes, let's make everyone happy

exactly (this market) Resolved YES 16 days ago

@Bayesian,

It’s not fair that you’re editing the description of a market that isn’t yours, especially when you hold a significant position in it.

With the updated description, it’s no longer clear that the author explicitly meant the exact release - just like in your other market about Gemini 3.0.

Yes, this market is basically just a duplicate.

😂