AI EconomicsJun 2024
Foundation model cost curves flatten, eval rigor becomes differentiator
- Insight
- Claude 3.5 Sonnet and GPT-4o mini pricing shifts mean evaluation coverage and prompt governance now drive ROI more than model selection.
- Signal
- Teams investing in eval harnesses are cutting re-run spend by 28%.
- Suggested action
- Prioritise evaluation infrastructure and prompt libraries over bespoke fine-tunes this quarter.