6 months from now will I judge that LLMs had already peaked by Nov 2024?
➕
Plus
23
Ṁ1963
2025
14%
chance

I will compare the state of the art LLMs in June 2026 with current LLMs. I will resolve YES if there's not a marked improvement, no credible, verifiable improvement on the horizon and my judgement is that LLMs 'peaked' around end of 2024. I will resolve NO if something like GPT-5 comes out and it's a (subjectively) step improvement from GPT-4. Improvements around the edges and stuff like multimodality don't count. If Google releases something new that catches up with Claude 3.5/GPT-4 that doesn't count.

I use LLMs extensively daily and my friends have described me as having "such great judgment" but this market is still vibes-based at the end of the day so caveat emptor. I will answer any/all relevant clarifying questions in the comments and add stuff to the description if it helps clear things up. While most of the resolution will be mine, I will allow some external consensus to influence me

I will not trade in the market to avoid (appearance of) conflicts of interest

Get
Ṁ1,000
and
S3.00
Sort by:

Do you have a working definition of "LLM" for these purposes? I see agent models gaining quickly on traditional LLMs, so a lot depends on which future models are considered LLMs, versus more complex systems of which an LLM is only one component.

is this gonna be resolved in June 2026 (18 months from now) or 6 months from now (June 2025)? its a no either way but clarification would be nice

@fakebaechallenge good catch. fixed it

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules