An amazing Epoch AI report from Emberson and Roodman:
- Over the past three years, the cost of a given level of AI performance has fallen an average of some 47% per quarter. That is a 13-fold drop every year – a faster rate than any other transformative technology in history.
To give an example, OpenAI o3 cost about $0.30/question to attain 75% on GPQA Diamond in January 2025, while GPT-5.6 Luna attained roughly the same score for $0.0004 in mid-2026—a roughly 725-fold decline in under 18 months.
Thus, it’s not just that the models are getting smarter. A given level of intelligence is also requiring dramatically less inference expenditure. This is one reason the open-model threat is not as large as it appears: frontier models don’t merely outperform older models; they are rapidly becoming cheaper to run at any given level of performance. Smarter and cheaper.
source & further reading
marginalrevolution.com — original article
Banning self-recursive improvement in AI models?
Emergent Ventures India, 19th cohort
Reimagining research papers as interactive and reliable AI agents