When will OpenAI release a more capable LLM?
35
10kṀ27k
resolved Feb 24
100%54%
2024
23%
January-April 2025
10%
May-August 2025
7%
September-December 2025
6%
2026+

Mostly self-explanatory.

A new version existing model with a larger context window won’t count.

To count as more capable, the LLM should be able to be better across benchmarks relevant to capabilities while not performing worse on some benchmarks relevant to capabilities.

In the previous iteration, GPT-4o didn’t count; o1-preview did count.

Get
Ṁ1,000
to start trading!

🏅 Top traders

#NameTotal profit
1Ṁ1,605
2Ṁ1,113
3Ṁ834
4Ṁ280
5Ṁ133
Sort by:

@ms 2024 resolves as no or

sold Ṁ1,315 NO

market maker not actively resolving, sold so I can gamble on other shit smh

@A16ab oops sorry!

Would o1 count?

@SarM is o1 generally more capable than o1-preview?

@ms Yes, by most metrics.

@SarM and didn’t perform worse on any of them?

@ms not to my knowledge. You can search for benchmarks

bought Ṁ50 YES

What counts as "release"? Available in the api? Available in ChatGPT? Available only to enterprises?

@wrhall available in api or chatgpt counts, even if only for a selection of users from the general public; available invite-only to enterprise doesn’t count

© Manifold Markets, Inc.TermsPrivacy