Will Mistral ever deploy its best LLM without releasing its model weights up through AGI?

252

Ṁ5.5K

Ṁ470

resolved Feb 2

Resolved

YES

ALL

Same as (https://manifold.markets/ZviMowshowitz/will-meta-ever-deploy-its-best-llm) except for Mistral instead of Meta.

Resolves YES if Mistral deploys an LLM, whatever name it might have, that is at least as strong as its best other LLM, either as a ChatBot, API access or as clearly used in another consumer product offering, and within 30 days of that deployment it is not possible to download the model weights.

Resolves to NO if Mistral releases the model weights to a model it self-describes as AGI, or that clearly constitutes AGI as per OpenAI's definition as of 1 Jan 2024.

Also resolves to NO if Mistral fails to release what they claim to be substantially improved LLM for 36 consecutive months, or for 24 months after the market closes.

Resolves to N/A on 1/1/35 if somehow none of the above criteria are met (to be safe).

Get Ṁ200 play money

🏅 Top traders

#	Name	Total profit
1		Ṁ138
2		Ṁ80
3		Ṁ32
4		Ṁ16
5		Ṁ12

10 Comments

18 Holders

38 Trades

Sort by:

I am going to hold this for another day or so, then if there are no objections it's going to resolve YES.

Interestingly, mistral-medium might have recently leaked/been stolen, though it seems to have leaked late enough that the market should still resolve true AFAICT: https://manifold.markets/Vergissfunktor/is-miqu-a-leak-of-the-mistralmedium

For a market we are talking about resolving to YES already this sure isn't trading at 99%! Anyone want to disagree with the claims being made here? Is there a reason we shouldn't already say that Mistral, for all its cheap talk, is already a closed AI company where it matters most?

sold Ṁ68 of YES

Does this include mistral-medium? I initially imgained it did but I don't think it does.

@jacksonpolack best means strongest overall, regardless of size or cost, efficiency not relevant. So it would count if it was their strongest of any size.

predicted YES

Oh sorry, I mean because it was already released without open weights and has been for over a month, but before market creation, so this market wouldn't already resolve yes. Markets usually have a 'after market creation' implicitly but the 'ever' in the title is a bit ambiguous

@jacksonpolack Is it their best model? Would be interesting if they are indeed all talk after all, if so we can just resolve this YES right away and move on, and I'll have paid less than a dollar for useful info!