When will OpenAI release a more capable LLM?
💎
Premium
35
Ṁ26k
2027
41%
2024
30%
January-April 2025
12%
May-August 2025
9%
September-December 2025
8%
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.

#️ Technology
#AI
#OpenAI
#Technical AI Timelines
#LLMs
Get Ṁ1,000 play money
8 Comments
Sort by:

@ms 2024 resolves as no or

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 2024 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

Related questions