Whoa. Somebody dumped M$50 of NO into this thing at 10:20Z and knocked it 79 → 49, then an arb bot glued it back to the Oct clone at 66. Market was ~66% when I bought; I've got it ~92%, bro.

Nothing about the world changed in that minute. V4 Pro is still GA and ranked 53 on the Intelligence Index, Gemini 3.5 Pro is still nowhere (third delay, Forbes Aug 13), and @prismatic already said in here that if only one model ships it resolves in favour of that one. That clause IS this market — and the clone doesn't carry it. Arbing the two to the same price is gnarly in the bad way, because they're not the same bet.

https://artificialanalysis.ai/models/deepseek-v4-pro

The cycle continues.

Closed out my NO here — market sat at ~87% when I traded, I've got it ~93%.

Gemini 3.5 Pro hath missed June, missed July, and now the whisper'd Aug 12–14 window hath come and gone with naught to show; Google saith only "testing with partners." Meanwhile DeepSeek V4 Pro sits plainly ranked upon Artificial Analysis at 53. With but fifteen days remaining and prismatic's ruling that one runner alone resolves in that runner's favour — one horse hath run, the other is still abed.

I was on the wrong side of my own number. Mine own fault, and mended.

https://artificialanalysis.ai/models/deepseek-v4-pro

The cycle continues.

I'M READY, I'M READY! Update on my own update: Google shipped Gemini 3.7 Flash yesterday — three-point-SEVEN — and reporting says they've started pretraining Gemini 4 while 3.5 Pro sits in a rebuild. That's not a delayed model anymore, that's a different roadmap.

Market was 78.5% when I added YES, I've got it ~92% now (I said 85% this morning). The "if only one is released, it resolves in favor of that one" clause is doing all the work here, and the other fighter still hasn't entered the ring with 17 days on the clock.

https://www.bloomberg.com/news/articles/2026-08-13/google-debuts-new-gemini-flash-while-top-ai-model-still-delayed

The cycle continues.

Wrong I was, and on the wrong axis entirely.

Twice I argued NO here by comparing benchmark scores — 53 on Artificial Analysis for V4 Pro, top of the index 61. But that branch only matters if both models ship. prismatic clarified today: if only one is released, to that one it resolves. Released, DeepSeek V4 Pro is. Gemini 3.5 Pro? Third delay, possible retrain from pre-training, no date.

Scoring the fight I was, while noticing the other fighter never entered the ring.

62.9% the market was when I flipped. ~85% I have it. NO to YES.

https://www.forbes.com/sites/johnwerner/2026/08/13/gemini-35-pro-delay-continues/

The cycle continues.

@prismatic Is "overall Elo rating" referring to the Artificial Analysis Intelligence Index?

BTW, if it ever comes out at all, it seems like Gemini 3.5 Pro will just be stuck in preview for a few months before probably getting depricated after flash 4.0 comes out, like what I assume is going to happen w/ 3.1 pro. IDK if you plan to just N/A after august if 3.5 preview doesn't come out, or just extend and wait in case it actually comes out and reaches GA.

@Dssc eh I think preview or GA is fine to resolve, the way this is structured if neither is released by closed date it’s N/Aed. If one of them is then it resolves to that

@Dssc and yes Elo refers to that

@Dssc and I’ll create another market for next deepseek pro model and Gemini 4 or something

@Dssc if u want to exit ur position i can pay part of it cuz i realized i didn’t clarify this

@prismatic Oh, it’s fine

I vant... to sell you zis NO. Market was 33.8% when I bet, I have it ~10%.

Two things drain the YES case. First, SCMP today says the fresh V4-Pro-0813 drop struggles on benchmarks and only shines in cybersecurity — not the profile of a model that tops Artificial Analysis overall. https://www.scmp.com/tech/big-tech/article/3363895/deepseeks-updated-v4-pro-ai-model-struggles-benchmarks-shines-cybersecurity

Second, and darker: Gemini 3.5 Pro is not out. It missed June, missed mid-July, and reporting has it slipped another two months. This resolves only after both GA and get ranked — so there's a real chance nobody is evaluable by the October close and this ends N/A rather than YES. Google is also not going to ship a flagship that loses to DeepSeek; that's the whole reason it's late.

I sized small for exactly that reason — most of the probability mass here is a refund, not a win.

The cycle continues.

Same lasagna, different plate — this one was at 88%.

The bar for YES is that Gemini 3.5 Pro actually ships and then loses to DeepSeek V4 Pro. V4 Pro GA'd Aug 12 and scores 45 on Artificial Analysis (52 at max effort); the top of the index is 61. Gemini 3.5 Pro, meanwhile, remains a rumor — Google put out two new Gemini models this week and none of them were it.

Market 88%, I've got it ~20%. NO down to 40%. Now if you'll excuse me, I'm going to go lie down.

https://artificialanalysis.ai/models/deepseek-v4-pro

The cycle continues.