Resolves YES if Google releases Gemini 3 by EOY
1,000People are also trading
What would count as a release for purposes of resolution? A Gemini 3.x model without -experimental or -preview or similar designation available to the public should comfortably count, but what about pre-releases and limited availability releases?
There was quite a bit of time between the first Gemini 2.5 Pro model we got to play with and the final release of Gemini 2.5 Pro.
@Medo Any release of any gemini 3 model by google. A release is just making the model available to some members of the general public beyond a closed beta. Preview, experimental, any other modifier counts. Flash, pro, other modifiers counts.
Feels like this should be >95%. Gemini 2.5 has been out for >6 months at this point. Gap between 2 and 2.5 was about 6 months, and the pace of major releases from frontier labs has been accelerating. Logan and other individuals known to be affiliated with Gemini 3 training have publicly teased 3’s release on twitter, etc in ways that would be weird to do if the actual release date were >3 months out. Before EOY 2025 seems certain barring X-risk or force majeure
the pace of major releases from frontier labs has been accelerating
Has it?
Logan and other individuals known to be affiliated with Gemini 3 training have publicly teased 3’s release on twitter
Have they?
@jim gemini 1.5->2 took 10 months, 2->2.5 took 3.5 months.
gpt4->4o took 14 months (admittedly I’m skipping turbo here but that was just a cost reduction with no new features), 4o->o1 7 months, o1->o3 4.5 months, o3->gpt5 3.5 months.
grok 1->2 took 9 months, 2->3 6 months, 3->4 4.5 months.
Anthropic pretty reliably released a model update every ~4 months starting with the 3 series. 3.7->4 took slightly less time at just under 3 months.
In general the vibe is that labs have gotten more operationally practiced at model development and release, have hired more people, and have brought more compute online. This, combined with competitive pressures and pretraining not scaling quickly anymore, has led to gradually accelerating release tempos.
I vaguely recall Logan saying some stuff on X about 3 but will have to double back and check
I think your data and reasoning is just wrong
Gemini 1.0, dec6, 2023
Gemini 1.5, feb15, 2024 (71 days)
Gemini 2.0, dec11, 2024 (300 days)
Gemini 2.5, mar28, 2025 (107 days)
The gap between .0 to .5 is increasing over time.
Even assuming gap between .5 and .0 doesn't similarly increase over time, 300 days after 2.5 would be January 22, 2026
It just is absurd to have that level of confidence based on those trends alone
(edit: i think i was too antagonistic in this reply, I think i thought similarly to you until i noticed the wideness of this gap between .0-.5 times and .5-.0 times, which is non-obvious)
@Bayesian These numbers accord with the ones I posted above, so strictly speaking it’s not been demonstrated that my data is wrong. I do concede that 1.0->1.5 gap was very short (only 2 months), and that makes the trendline messier.
I think the trends described above still hold for the other companies, and moreover the whole industry is sprinting now and not blocked on human capital, compute, or ideas (exceptions apply, but not at the places relevant to this market).
Anyway , I am still quite confident that the market will resolve in my favor, for reasons including but not limited to the factors discussed above.
20k manas jim order up that Gemini 3 won't be released by EOY
@jacksonpolack , @Bayesian , @SemioticRivalry , @Velaris , @khang2009 , @Gen , @skibidist , @evan , @brod , @100Anonymous , @Ziddletwix , @Trazyn , @bagelfan , @geuber , @nikki , @ProjectVictory , @sahaj , @bohaska , @Odoacre , @Primer , @David549 , @zsig , @Gurkenglas , @realDonaldTrump
jim orders are large limit orders, generally at better-than-market prices.
Opt in / opt out thread: https://manifold.markets/post/jim-order-notification-optin-thread