Will there be evidence in 2025 that in April 2023, OpenAI had a GPT-4.5 or higher model?
➕
Plus
20
Ṁ425
2026
16%
chance

Resolves YES if in 2025 there is a lot of evidence that OpenAI had a base model with a LM language-modelling loss lower than a model trained with 7X the compute using the GPT-4 scaling law.

This excludes retrieval and other "cheap to apply to GPT-4, known capability enhancements".

A "partially trained" model with a sufficiently low loss counts.

See https://twitter.com/taoroalin/status/1647049143519694848

Get
Ṁ1,000
and
S1.00
Sort by:

OpenAI has already said they aren't training GPT-5, and they said they would release GPT-5 incrementally.